Ontology-based representation and reasoning about the history of science more

by Ilaria Corda

Master by Research (Mres) Thesis, School of Computing, University of Leeds, 2007

Event Modeling, Ontology Engineering, History Ontologies, Ontology Methodology, and Time Modelling

Ontology-based Representation and Reasoning about the History of Science by Ilaria Corda Submitted in accordance with the requirements for the degree of Master of Science by Research. The University of Leeds School of Computing April 2007 The candidate conﬁrms that the work submitted is his own and that the appropriate credit has been given where reference has been made to the work of others. This copy has been supplied on the understanding that it is copyright material and that no quotation from the thesis may be published without proper acknowledgement. Abstract The use of ontologies enables semantically enriched access to a variety of digital resources. Historical domains impose a number of challenges to creating ontologies, e.g. modeling temporal relations, handling subjectivity, and dealing with vagueness. This research develops a case study in the History of Science that illustrates how to conceptualise and reason about a historical domain, and suggests an approach to modeling time and temporal relations. Based on existing methodologies for ontology construction, a methodology for conceptualising (a part of) the History of Science domain has been derived taking into account that the author has acted as both a domain expert and a knowledge engineer. Following the methodology, main concepts and relations in the History of Science have been identiﬁed. Special attention is paid at modeling temporal concepts and relations. A framework for conceptualising and reasoning about the History of Science is presented, combining Davidson’s theory of events to represent temporal categories and Allen’s interval logic to reason about temporal relations. i Acknowledgements I would like to express my gratitude to Dr Vania Dimitrova and Dr. Brandon Bennett for their patience and encouragement that carried me on through difﬁcult times and for their insights and suggestions that helped to shape my research skills. Special thanks to Dr.Vania Dimitrova during the ﬁnal stage of my Msc studies and write up. Without her valuable feedback, this dissertation would not have been possible. My special thanks are due to Prof. Giancarlo Nonnoi, my B.A. supervisor, for having persuaded and helped me to study abroad. I will never thank him enough for what he did for me. Many Thanks to my mother, my Italian friends and all people who believe in me. I gratefully acknowledge the funding sources that made my Msc work possible. ii Conventions In this thesis, the terms knowledge engineer and ontologist are assumed to have the same meaning; referring to a person who gathers the knowledge of the expert and translates into machine-readable representation. The terms domain expert and expert are also assumed equal; referring to a person with special knowledge or skills in a particular area of expertise. For the convenience of thesis writing, the pronoun we and the adjective our refer to the author. iii Contents 1 2 Introduction Background and related work 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Modeling challenges in historical domains . . . . . . . . . . . . . 2.3 History of Science domain and challenges addressed in this thesis 2.4 Relevant methodologies for ontology development . . . . . . . . 2.4.1 METHONTOLOGY . . . . . . . . . . . . . . . . . . . . 2.4.2 Gruninger and Fox’s methodology . . . . . . . . . . . . . 2.4.3 Uschold and King’s method . . . . . . . . . . . . . . . . 2.4.4 Ordnance Survey’s methodology . . . . . . . . . . . . . . 2.5 Modeling time . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5.1 Formal representations of temporal information . . . . . . 2.5.2 Davidson’s theory of events . . . . . . . . . . . . . . . . 2.5.3 Allen’s Interval algebra . . . . . . . . . . . . . . . . . . . 2.5.4 Time in upper ontologies . . . . . . . . . . . . . . . . . . 2.5.5 Upper ontologies for temporal concepts . . . . . . . . . . 2.5.6 Modeling time in History ontologies . . . . . . . . . . . . 2.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Building a History of Science Ontology: Main Phases 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . 3.2 Pre - conceptualization . . . . . . . . . . . . . . . 3.3 Conceptualization . . . . . . . . . . . . . . . . . . 3.4 Logical representation and coding . . . . . . . . . 3.5 Summary . . . . . . . . . . . . . . . . . . . . . . 1 6 6 6 8 9 9 12 14 16 19 19 21 22 24 27 30 35 37 37 38 41 43 48 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv 4 Modeling Time in a History of Science Ontology 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . 4.2 Time concepts in History of Science ontology . . . 4.3 Modeling time in relations: initial approach . . . . 4.4 Modeling time in relations: use of Davidson events 4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 50 51 54 57 62 63 63 64 64 68 70 74 75 76 78 79 80 80 81 81 81 82 84 84 85 85 87 89 5 Querying the History of Science Ontology 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Ontology-based query modes . . . . . . . . . . . . . . . 5.2.1 Concept mode . . . . . . . . . . . . . . . . . . 5.2.2 Relation mode . . . . . . . . . . . . . . . . . . 5.2.3 Time-Event mode . . . . . . . . . . . . . . . . . 5.3 Domain-speciﬁc questions . . . . . . . . . . . . . . . . 5.3.1 Who questions . . . . . . . . . . . . . . . . . . 5.3.2 What questions . . . . . . . . . . . . . . . . . . 5.3.3 Where questions . . . . . . . . . . . . . . . . . 5.3.4 When questions . . . . . . . . . . . . . . . . . . 5.3.5 Combined Type questions . . . . . . . . . . . . 5.4 Using queries to verify and expand the ontology . . . . . 5.4.1 Syntactical errors . . . . . . . . . . . . . . . . . 5.4.2 Hierarchical inconsistencies and redundancies . . 5.4.3 Population of the ontology for answer the queries 5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions and Future work 6.1 Synopsis . . . . . . . . . . . . . 6.2 Main contributions of this thesis 6.3 Limitations . . . . . . . . . . . 6.3.1 Future work . . . . . . . Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A Expanded queries 97 A.1 Example queries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 A.1.1 Concept Mode queries: additional examples . . . . . . . . . . . . 97 A.1.2 Relation Mode queries: additional examples . . . . . . . . . . . . 109 v B Ontology statistics and glossary entries B.1 Statistics of ontology size . . . . . . . . . . . . . . . B.2 Glossary entries: upper classes . . . . . . . . . . . . B.3 Glossary entries: lower classes . . . . . . . . . . . . B.3.1 Person: lower classes . . . . . . . . . . . . . B.3.1.1 Scientist: subclasses . . . . . . . B.3.1.2 Philosopher: subclasses . . . . . . B.3.2 Role: lower classes . . . . . . . . . . . . . . B.3.3 Place: lower classes . . . . . . . . . . . . . B.3.3.1 Geopolitical Area: subclasses . . . B.3.3.2 Astro Area: subclasses . . . . . . B.3.4 Field of study: lower classes . . . . . . . . . B.3.5 Belief: lower classes . . . . . . . . . . . . . B.3.6 Document: lower classes . . . . . . . . . . . B.3.7 Mode of reasoning class: lower classes . . . B.3.8 Doctrine: lower classes . . . . . . . . . . . . B.3.9 Event: lower classes . . . . . . . . . . . . . B.3.9.1 Observation: subclasses . . . . . . B.3.10 Philosophical doctrine class: subclasses . . . B.3.11 Method: lower classes . . . . . . . . . . . . B.3.12 Model: lower classes . . . . . . . . . . . . . B.3.13 Group of people: lower classes . . . . . . . . B.3.13.1 Academic organization: subclasses B.3.14 Time: lower classes . . . . . . . . . . . . . . B.3.14.1 Historical period:subclasses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 112 112 113 113 114 114 115 115 115 116 116 116 117 117 118 118 118 119 119 120 120 120 121 121 vi List of Figures 2.1 2.2 2.3 2.4 2.5 Temporal categories in the Cyc’s class taxonomy. Primitives in Sowa ontology. . . . . . . . . . . . Sowa’s Process type. . . . . . . . . . . . . . . . OWL time ontology: hierarchy . . . . . . . . . . KSL ontology’s hierarchy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 26 27 28 29 vii List of Tables 2.1 2.2 2.3 Glossary of term in METHONTOLOGY: an example in History of Science 10 Binary Relation Table: an example in History of Science . . . . . . . . . 11 Examples of Core and Secondary concepts from History of Science Ontology, deﬁned according to the Ordnance Survey’s Methodology . . . . . 17 History of Science Glossary of Terms: Example Entries . . History of Science Relations Table: Example Entries . . . History of Science Instances Table: Example Entries . . . History of Science Event Relations Table: Example entries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 43 43 43 112 112 113 113 114 114 115 115 115 116 116 116 117 117 118 118 118 119 3.1 3.2 3.3 3.4 B.1 Overview of the ontology size . . . . . . . . . . . . . . . . . . . . . . . B.2 Statistics on the ontology size: transitive, inverse and symmetrical upper relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B.3 Glossary entities for upper classes . . . . . . . . . . . . . . . . . . . . . B.4 Person: lower classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . B.5 Scientist: subclasses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B.6 Philosopher: subclasses . . . . . . . . . . . . . . . . . . . . . . . . . . . B.7 Role: lower classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B.8 Place: lower classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B.9 Geopolitical Area: subclasses . . . . . . . . . . . . . . . . . . . . . . . . B.10 Astro area: subclasses . . . . . . . . . . . . . . . . . . . . . . . . . . . . B.11 Field of study: lower classes . . . . . . . . . . . . . . . . . . . . . . . . B.12 Belief: lower classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . B.13 Document: lower classes . . . . . . . . . . . . . . . . . . . . . . . . . . B.14 Reasoning: lower classes . . . . . . . . . . . . . . . . . . . . . . . . . . B.15 Doctrine: lower classes . . . . . . . . . . . . . . . . . . . . . . . . . . . B.16 Event: lower classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B.17 Observation: subclasses . . . . . . . . . . . . . . . . . . . . . . . . . . . B.18 Philosophical doctrine: subclasses . . . . . . . . . . . . . . . . . . . . . viii B.19 B.20 B.21 B.22 B.23 B.24 Method: lower classes . . . . . . Model: lower classes . . . . . . . Group of people: lower classes . . Academic organization: subclasses Time: lower classes . . . . . . . . Historical period: subclasses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 120 120 120 121 121 ix Chapter 1 Introduction Over the past years the World Wide Web has became a widely-consulted source of information for an increasing number of users. Resources, such as Wikipedia,1 citeUlike,2 and Perseus,3 show the abundance of information available on line. Despite the existence of advanced search engines (e.g Google or Yahoo), people are still facing the problem of ﬁnding the appropriate digital resources when searching through unstructured data on the web. The use of ontologies can enable semantically enriched access to a variety of digital resources. We follow the view that using ontological structures for representing and reasoning about a domain can contribute to the development of intelligent technologies that empower people in gaining a better understanding of the domain and provide more effective ways for discovering digital resources. With this in mind, we will conduct a case study to represent and reason about a speciﬁc domain - the History of Science. In this thesis, the notion of ontology suggested by Gruber and the more recent interpretation proposed by Guarino will be followed. According to Gruber, an ontology is an explicit speciﬁcation of a conceptualization [41], 1 Wikipedia 2 citeUlike is an online collective encyclopedia: http://www.wikipedia.org/ (Visited, April 2006 is a community resource for sharing personal references: http://www.citeulike.org/ (Vis- ited, March 2006) 3 Perseus is an online digital library offering access to public libraries all over the world: http://www.perseus.tufts.edu/ (Visited, March 2006) 1 Chapter 1 Introduction 2 while Guarino advocates that ontologies are based on logical conventions, and deﬁnes an ontology as a set of logical axioms designed to account for the intended meaning of a vocabulary [44]. Domain ontologies provide vocabularies about concepts and their relationships within a particular domain (medical, historical, enterprise, etc). Such domain-dependent knowledge represents the specialization of concepts which are already deﬁned in top level ontologies [38]. History ontologies are domain-speciﬁc ontologies aimed at representing the semantic structure of historical domains by identifying an underlying set of relations between concepts and modeling them with appropriate formal speciﬁcations. History is a complex and loosely structured domain which imposes a number of challenges to creating ontologies such as modeling temporal relations, handling subjectivity, and dealing with vagueness and uncertainty. The main goal of this research is conceptualise (part of) the History of Science by focusing on modeling time and reasoning about time dependencies. To achieve this goal, we will derive an appropriate methodology for ontology development and will apply this methodology to conceptualize our domain. We will then develop a systematic framework for representing and reasoning on the domain of History of Science, focusing on time dependencies between History of Science events. With the goal in mind, the following research questions will be addressed in this thesis: • Which methodological principles can be followed for the developing of a History of Science ontology? What degree of involvement should the domain expert have? How should a methodology be scaled to comply with the time constraints and person limitations or case when both roles, domain expert and knowledge engineer, are performed by the same person? • Can ontological structures appropriately represent temporal speciﬁcation in History of Science? What time concepts should be integrated in a History of Science ontology? How can time speciﬁcations be added to relations and descriptions of events? • How can an ontology-based temporal representation efﬁciently support reasoning on the domain? What reasoning mechanism can be implemented to reason about ontological relations? How can abstract reasoning rules be utilised in domain speciﬁc rules to infer knowledge about the History of Science? Chapter 1 Introduction 3 The work resulted in the following achievements: • Development of a three phase methodology for building a History ontology: The design of a speciﬁc methodology enabled us to conceptualise the domain, by considering the main activities and tasks to be performed. The role of the domain expert covers all aspects of the ontology building, as in this case study the expert acted as knowledge engineer too. • Development of a framework for adding temporal speciﬁcation: We have chosen to address time modeling as crucial aspect of any historical domain. We developed an elaborated framework for dealing with time inclusion and representation of History of Science events, by following Davidson’s theory of events. • Development of a corpus of rules for reasoning on the domain: Querying and reasoning on the domain have been addressed by considering abstract and domain oriented rules. The former are applicable to reasoning about historical domains, while the latter were tailored to the speciﬁc historical domain considered here and were used for verifying the consistency and expressiveness of the History of Science ontology. Motivating scenario To illustrate the need for the ontology that will be developed in this thesis, we will outline a possible scenario of using an ontology-based representation to facilitate searching through digital resources (e.g. web resources) in the History of Science. Consider a situation where a learner has to explore materials from a digital collection in order to accomplish a learning goal (e.g. to write an essay on a topic related to the History of Science). Assume that the digital resources are enhanced with metadata linked to an ontology that describes the conceptual relations within the domain. We will refer to the topic of Astronomical Revolution encompassing concepts required for this assignment. Consider Eric, to be an A level student who has an assignment, for which he needs to describe the key achievements during the Astronomical Revolution, including: • Main scientiﬁc models: Theories and Laws for explaining phenomena and their interdependence (e.g. Which theories were deﬁned during the Astronomical revolution? When did this happen? How were these theories extended or replaced by other theories?). Chapter 1 Introduction 4 • Main scientiﬁc events:. Discoveries and inventions holding speciﬁc time-space locations (e.g. Which discoveries and inventions happened during the Astronomical Revolution? Who made them and where? Which events preceded or succeeded particular discoveries or inventions?). • Major contributors and their research activities: Scientists, their related ﬁelds of study, major achievements and publications (e.g. Which scientists worked during the Astronomical Revolution? What ﬁeld did they work in? What did they write? What did they publish? When and where did it happen?). • Social network relationships: Collaborations and/or inﬂuences between scientists who worked on similar issues (e.g Who collaborated with whom? What group did a given scientist work with? How were scientists inﬂuenced by others?). Eric does not have any previous knowledge about this topic and he might not be able to provide speciﬁc key words to a search engine (e.g. when searching for Kepler Eric will not be able to specify Brahe as well since Eric will not know that both scientists collaborated and that Brahe inﬂuenced Kepler). Eric is likely to get a large quantity of unstructured results and will be unable to ﬁnd what is relevant to his essay. He can then feel confused and frustrated. To remedy this, a History of Science ontology can be used in two ways. Firstly, Eric may ask an intelligent assistant to help him ﬁnding answers on the above questions, so Eric can derive more speciﬁc key words that will lead to more appropriate resources. On the other hand, a smart search algorithm can make use of the ontology to expand or to clarify Eric’s set of keywords (e.g. when typing ’influence’ and ’Kepler’, the smart search can add Brahe in the set of search terms). This will lead to more efﬁciently supporting task-oriented navigation, avoiding time consuming searches and information overload, and enhancing the users’ experiences with digital resources. This scenario will be used for deﬁning the scope of our ontology. Thesis overview This thesis is organized in ﬁve parts: Background and related work: In Chapter 2, we will discuss the characteristics of the domain (History of Science) together with main requirements and modeling challenges Chapter 1 Introduction 5 in History ontologies. We will also review relevant methodologies for developing ontologies, as well as theoretical and ontological approaches for dealing with temporal information. This review will provide the basis for the development of our ontology construction methodology and for deﬁning a time modeling framework. Methodology: The second part, Chapter 3, together with chapters 4 and 5, form the core of this thesis. In Chapter 3, we will propose our three-phase methodology for building a History of Science ontology. We will examine the characteristics of each phase (pre-conceptualization, conceptualization, logical and encoding) and will discuss the role of the domain expert throughout the entire process. Time modeling: In Chapter 4, we will present our approach for modeling time categories and relations, by considering three main dimensions in which time can be expressed: time concepts, temporal relations, and instantiated relations corresponding to events. We will illustrate the advantages of using the Davidson’s theory of events for systematically adding time in historical domains. We will also propose the use of type-token distinction for identifying relations in which time is required to be added. Querying and reasoning about the domain: In Chapter 5, we will present rules for reasoning about the History of Science, by considering three ontology-based modes across our domain: Concept-based, Relation-based, and Time-Event-based. We will exemplify these modes by introducing a set of questions used for writing domain-oriented rules. Finally, we will discuss the use of these questions in verifying the consistency and checking the expressiveness of our ontology. Conclusion and Future work: The last chapter will summarise the main aspects of this research, pointing at the main contributions of this thesis. We will also outline the limitations of our approach and will sketch out directions for future enhancements and extensions. Chapter 2 Background and related work 2.1 Introduction In this chapter, we will give an account of the background reading we have considered in our work. Firstly, we will discuss the characteristics and modeling challenges in historical domains (Section 2.2) which will be appropriately considered according to our case study: a History of Science ontology. In Section 2.3 we will outline the challenges addressed in this work, focusing on representing and reasoning on temporal dimensions. Section 2.4 will review a number of existing methodologies for the sake of developing a domain-oriented methodology for our case study. Each methodology was considered according to the relevance of its features for our domain. Section 2.5 will consider theoretical approaches for dealing with time, based on key contributions from Philosophy and AI. In Sections 2.5.2 and 2.5.3 we will illustrate the two approaches which have been integrated in our framework, as described in Chapter 4 and 5: Davidson’s theory of events and Allen’s interval algebra. Finally, time modeling approaches in ontologies will be reviewed according to different granularities of temporal representation. 2.2 Modeling challenges in historical domains History is a complex, loosely structured domain. Semantic representations of historical facts have to accommodate the characteristics of the History domain, which brings spe- 6 Chapter 2 Background and related work 7 ciﬁc modeling challenges. As summarized in [72] [73], the following main issues have to be considered when building a history ontology: • General coverage of the History domain: The ontology should encompass general concepts related to History, while at the same time allowing to decompose the knowledge into specialized concepts and relations, as required. • Modeling time dependency and temporal changes over time: Historical facts are temporally interdependent. Furthermore, historical events depend on the social and political context in which they took place. An ontological model of the History domain should adopt an appropriate framework for representing and reasoning about time speciﬁcations. • Dealing with vagueness in historical descriptions: Vagueness is manifested by inability to have precise boundaries of historical periods or precise time measurement of events. For example, there are no crisp boundaries of ‘Early modern’ in Europe. Often, time speciﬁcations are expressed following different level of granularity such as year, century, or statements with imprecise time. The mechanisms to represent and reason upon vague descriptions usually employs fuzzy logic or supervaluation. • Dealing with uncertainty about historical facts: Uncertainty relates to facts based on missing or partial sources, which can lead to dealing with unknown information (e.g. dates, people, places) or contradictory facts. For instance, the KantLaplace theory was developed independently by the two authors who lived in different places. The scientiﬁc community can not reach a consensus about who formulated this ﬁrst, as there is no conclusive information. Similarly, the principle of conservation of energy was formulated by twelve scientists between 1830 and 1850. It cannot be said for certain who ﬁrst discovered that principle. • Accommodating subjectivity in historical knowledge: Subjectivity, which is directly linked to uncertainty, is related to different opinions or multiple interpretations of the same events. There are many cases when the historical community holds diverse and sometime opposite perspectives when interpreting historical events. Possible ways to address subjectivity can be based on representing and reasoning about context. In this thesis, we will consider a case study for building an ontology of a speciﬁc historical domain, focusing on one of the above challenges-modeling and reasoning about temporal dimensions. In the next section, we will present the nature of our domain of Chapter 2 Background and related work 8 interest-History of Science-and will justify the importance of modeling time in this domain. 2.3 History of Science domain and challenges addressed in this thesis The History of Science is a subﬁeld of History aiming at studying how people’s understanding of science has been changed over time. History of Science is considered to be a ﬁeld of research which deals with the expression, preservation, and change of human ideas over time. The History of Science studies the intellectual paths covering periods of time during which science developed systematic theories, and earlier theories were abandoned. Such domain is the result of many interdisciplinary inﬂuences from Politics, Sociology, Psychology, History of ideas, as more than one aspect has to be considered when studying how the human understanding of scientiﬁc phenomena is subjected to continual changes. Among those, Porter and Teich emphasize that it is important to: evaluate the role of particular and disparate national and cultural traditions of thinking and mental work, the patterns of education, the channel of intellectual communication the opportunities for, or restriction upon, free thought and expression that operated within discrete language groups and under distinctive political jurisdictions [78]. Also, the History of Science describes how scientiﬁc development affected people’s life in general. For instance, the ‘Astronomical Revolution’ had not only a great impact on the image of the science, but it also represented one of the biggest conﬂicts between the Roman Catholic Church and the Aristotelian theory of the Universe. The nature of the History of Science points out that modeling time dependency and temporal changes over time is paramount in this domain. Knowing when an event happened, what happened at the same time, and what events preceded or succeeded the event, is critical for understanding the historical aspects of scientiﬁc developments. There are numerous examples of the need for modeling time in the History of Science. For instance, Galileo, called the ‘father of modern astronomy’, made a signiﬁcant breakthrough from the Aristotelian doctrine during the period in which Roman Catholic Church was the symbol of the authority against freedom of thought. Similarly, the Second World War had a great impact on the scientiﬁc community, causing enormous improvements in new sciences such as computer science and cybernetics. For example, during the World Chapter 2 Background and related work 9 War II Polish people became the most qualiﬁed at code-breaking in order to defend their unfavorable geographical position during the conﬂicts. We will present more examples in Chapters 4 and 5 of this thesis related to the scenario and the topic of the ontology developed in this thesis. Modeling time is the key issue to be addressed when building a History of Science ontology, which may precede addressing other challenges (i.e. vagueness, uncertainty and subjectivity). We have therefore chosen to address this challenge in the History of Science ontology developed in this thesis. Speciﬁcally, we will consider how to model and reason about time dependency between events, as well as how to discover relations within the scientiﬁc community, such as collaboration and inﬂuences. 2.4 Relevant methodologies for ontology development In this this section we will review a set of methodologies and methods for building ontologies, which can be relevant to our ontology development process, and have been deployed for the development of domain ontologies that comply with the design criteria identiﬁed by Gruber [41] [42]: clarity, coherence, extendibility, and minimal encoding bias and ontological commitment. It is important to note that the terms methodology and method have often been used interchangeably in the context of ontology construction [37] [18]. In this thesis, we will follow the IEEE deﬁnition [50] that a methodology is ‘a comprehensive, integrated series of techniques or methods creating a general systems theory of how class of thought-intensive work ought be performed’, while a method is ‘an orderly process or procedure used in the engineering of a product or performing a service’. Our review will include several methodologies and one method for ontology construction. Illustrations from the History of Science will be provided, when appropriate. 2.4.1 METHONTOLOGY METHONTOLOGY is considered as one of the most outstanding methodologies for building ontologies. It [39] has been used to develop several ontologies such as CHEMICAL [26], the Environmental pollutants ontologies [40] and Reference-Ontology [23]. Existing ontology development tools, such as Protege [24] and Ontoedit [27], can be applied for building ontologies following this methodology. METHONTOLOGY is based on the IEEE 1074-1995 description of software development life cycle [51], and identiﬁes a set of activities which an ontology performs during its lifetime [30]. Following the notion of evolving prototypes, METHONTOLOGY allows one to modify, remove, or add Chapter 2 Background and related work 10 terms in each version of the ontology. Each ontology version is considered as a prototype which is scheduled for performing a set of tasks by considering an expected time of completion, resources and arrangements. An ontology speciﬁcation activity is performed ﬁrst to describe the ontology’s purpose, scope, granularity, and sources used (e.g. interviews with experts, books and references) [26]. Once most of the knowledge has been acquired, the developers have to organize the collected data. The conceptualization activity aims to structure domain knowledge making use of representation models that are independent from any ontology language and tool [26]. The conceptualization activity converts an informal view of the domain into a semi-formal speciﬁcation using so called intermediate representations easily understandable by both ontologists and domain experts [37]. In this way, domain experts can play a crucial role in organizing the required knowledge without being discouraged by the language used to developed it. The ontologists, instead, can gradually move from the knowledge level to a machine-understandable level of formality. The conceptualization phase consists of several tasks that identify the main components of the ontology and integrate new terms within the existing representation. Task 1 is the building of a glossary of terms including all relevant terms with regard to the domain of interest, i.e. from upper concepts to instances, properties, etc. Table 2.1 illustrates glossary entries from the History of Science following METHONTOLOGY: Name Phenomenon Synonyms – Acronyms – Description Any state or process known through the senses rather than by intuition or reasoning The force of attraction between all masses in the universe Type Concept Gravity – – Concept Table 2.1: Glossary of term in METHONTOLOGY: an example in History of Science Task 2 refers to the construction of concept taxonomies. Once the glossary of terms contains a good number of entries, the ontologists can develop a concept classiﬁcationtree in which they state relations such as Subclass-of, Subclass-partition-of, Exhaustive-subclass-of etc. The next ﬁve steps (tasks 3-8) represent the core part of the ontology building and are closely linked to each other. Task 3 concerns the deﬁnition of a binary relations diagram between the concepts contained in the taxonomy. Once the relations-diagram has been generated, it is important to build a concept dictionary including all domain concepts and related instances, class and instance attributes for each concept, and, if appropriate, syno nyms and acronyms. Then, the ontologists have to detail the binary relations, instances and class attributes, organizing them in tables. Table 2.2 illustrates a binary relation from Chapter 2 Background and related work 11 the History of Science domain following METHONTOLOGY: Relation name Explain Source concept Model Source Card. N Target Concept Phenomenon Mathematical properties Transitive Inverse Relation is explained by Table 2.2: Binary Relation Table: an example in History of Science Tasks 9 and 10 are respectively the descriptions of the axioms and rules in the domain. They are performed by ontologists who describe the axioms needed in the ontology making use of tables which include the following information: (a) Axiom name, (b) Natural language description and logical formalization using ﬁrst order logic, (c) Concept, (d) referred attributes and (e) ad hoc relations and variables in use. Although the conceptualization tasks in METHONTOLOGY are not sequenced, GomezPerez et al. argue that some order needs to be followed to ensure the criteria of consistency and completeness [37]. For instance, if we decide to include an entry named discovery, it is more consistent to describe it ﬁrst in a glossary of terms specifying its natural language description than getting started from the axiom or rule layer. Following METHONTOLOGY, the knowledge acquisition activity occurs in three stages [26]: (a) preliminary meetings with experts to look at coarse grained knowledge, (b) in-depth study of the documentation as the domain experts should not spend time instructing the ontologists about the domain, and (c) knowledge specialization process by applying a top-down strategy (from the most general into more speciﬁc concepts). Several techniques for knowledge acquisition have been suggested like unstructured and structured interviews with experts and informal and formal text analysis. The former can be used both to sketch a draft of a requirements-speciﬁcation document and to gain detailed knowledge about concepts, their relations, attributes, etc. The latter involves either the analysis of the main concepts from recommended literature or non-automatic information extraction process which considers the recurrent patterns in the text.1 METHONTOLOGY considers the ontology evaluation as a key activity throughout the entire ontology development and recommends regular reviews by domain experts. METHONTOLOGY has been followed in large scale ontology development over long time periods involving teams of domain experts and ontologists who produce comprehensive and practically deployable ontologies. The scope of our project is much smaller, and the rigor of METHONTOLOGY activities and tasks does not appear applicable (e.g. it would not be realistic to have strict deﬁnitions of different ontology versions or to clearly is important to say that this technique does not involve any computational environment. Once recurrent structures are detected, (e.g. A is B), they are analyzed to derive natural language deﬁnition, attributes, etc [26] 1 It Chapter 2 Background and related work 12 distinguish the tasks for domain experts and ontologists). However, the importance of considering constant evolution of the ontology and performing evaluation throughout the whole ontology building process (especially during the conceptualization phase) will be taken into account in the methodology followed in our work, (see Chapter 3). 2.4.2 Gruninger and Fox’s methodology Gruninger and Fox’s methodology has been used in the TOVE project (TOronto Virtual Enterprise)2 [31] for building ontologies within the domain of business process and enterprise modeling. It considers the use of ﬁrst order logic for developing knowledge base systems in which the logical model is speciﬁed throughout a number of steps. The initial step is describing a motivating scenario that illustrates the use of the ontology [29]. The motivating scenario can be seen as a story telling activity pointing at situations and problems which have not been adequately addressed by any existing ontology [43]. This preliminary process is not only vital for informally deﬁning the scope of the ontology, but it can also point at candidate ontologies to be partially re-used. It can also provide an informal semantics for the terms and relations that will later be formally encoded. The next process is the formulation of informal competency questions. They are natural language questions grounded on the scenario and expressing the requirements the ontology needs to meet. Gruninger and Fox stress the importance of the competency questions during the whole development process, in order to evaluate and control the expressiveness of the ontology. These questions are not yet expressed in any formal language, however they can be further specialized as formal questions by using of the logical language of the ontology. Ideally, an ontology must be able to represent all these questions using its terminology, and articulating their answers through axioms and deﬁnitions [29]. According to Gruninger and Fox [43], the competency questions should not to be considered as simple queries. In fact, each question has to be deﬁned in a stratiﬁed manner in which higher level questions require the solution of low level questions by means of decomposition and composition operations. Thus, they can be divided into more atomic questions whose answer is used for more complex questions too. For example, a competency question related to a History of Science ontology can be: When, where and by whom was a particular phenomenon observed? The same competency question can be further decomposed into atomic questions: ontology project: http://www.eil.utoronto.ca/enterprise-modelling/tove/index.html (Visited, September 2006) 2 TOVE Chapter 2 Background and related work 13 When was a given phenomenon observed? Where was a given phenomenon observed? Who has observed a given phenomenon? After the informal competency questions are identiﬁed, they can be used for specifying the terminology vocabulary which is the third process of ontology development. From the questions, the ontologist will get a set of terms which will serve as the basis for specifying the terminology in a formal language. Then, from the answers s(he) will obtain the knowledge which will be included in the formal deﬁnition of concepts, relations between them, and axioms. The building of the formal apparatus of the ontology includes identifying objects and predicates in the domain of discourse [43]. The former are instances populating the ontology. The latter are unary and binary predicates used for representing concepts/attributes and binary relations, respectively, as well as n-ary predicates for expressing relations among objects. The following examples from the History of Science illustrate Gruninger and Fox’s terminology vocabulary. The notation used deﬁnes a class by using the predicate define-class, and assuming that anything after the question mark is a variable: define-class phenomenon(?phenomenon) define-class scientist(?scientist) scientist-name(?scientist?string) phenomenon-name(?phenomenon?string) observe(?scientist?phenomenon) The fourth process in Gruninger and Fox’s methodology is formally representing the competency questions in ﬁrst-order logic. The authors advocate that a proposal of extending or cleaning up an ontology must be associated to a set of formal competency questions since this is the only way to adequately evaluate an ontology throughout its entire life cycle [43]. The following example illustrates how an atomic competency questions shown above can be written in a ﬁrst order logic formalism: (∃x, y)(phenomenon(x) ∧ scientist(y) ∧ observe(y, x)) The next process includes specifying axioms using the chosen formalism. Axioms in the ontology are considered as the deﬁnitions of terms and pose constraints on their interpretation. They have the form of ﬁrst-order sentences which include the predicates in the ontology. Gruninger and Fox point out that specifying axioms represents the most challenging and difﬁcult aspect of ontology construction [43]. There is a reciprocal implication linking formal competency questions and axioms. Axioms are crucial for adequately Chapter 2 Background and related work 14 expressing the competency questions and characterizing their solutions. If the stated axioms are not sufﬁcient to represent the formal competency questions and describe their solutions, new axioms have to be added. The last process is establishing circumstances under which completeness can be achieved. It consists of deﬁning formally the conditions under which the solutions of the competency questions are complete. In the above example, we might not need any formulation of completeness since we are only required to list all scientists who meet our request. Although there are concerns that the Gruninger and Fox’s methodology does not identify the ontology life cycle and provides insufﬁcient detail about the recommended techniques and activities (e.g. those concerned with the formulating of competency questions) [29], this methodology introduces crucial processes related to the ontology development and applicable to our case study. Speciﬁcally, it advocates the use of a motivating scenario to deﬁne the scope of the ontology and derive competency questions, it stresses the importance of employing the competency questions to validate the ontology, and illustrates that logical representations can be used for encoding the domain knowledge. These issues will be considered in the methodology used in this thesis (see Chapter 3). 2.4.3 Uschold and King’s method Uschold and King’s method has been used for developing an enterprise modeling processes ontology, called Enterprise Ontology.3 Uschold and King’s method consists of four main processes: (a) Identifying the scope, (b) Building the ontology, (c) Evaluating the ontology, and (d) Documenting the ontology. Ontology development starts with the identiﬁcation of the ontology scope. Uschold and King suggest to identify a range of intended users in order to exemplify the scenario in which the ontology will be applied. The role of competency questions is also related to identifying the scope of the ontology. However, unlike Gruninger and Fox’s methodology which considers the identiﬁcation of competency questions to be crucial before developing an ontology, Uschold and King do not strictly place them prior to the conceptualization phase. The core process is the building of the ontology, which includes identifying and acquiring the domain knowledge (capture), expressing concepts and relations in a formal language (coding), and, if appropriate, reusing existing ontologies (integrating). The capture activity includes identifying concepts and relations by producing unambiguous textual deﬁnitions, recognizing instances which refer to concepts and relationships, and Ontology web page: http://www.aiai.ed.ac.uk/project/enterprise/enterprise/ontology.html (Visited, April 2006). 3 Enterprise Chapter 2 Background and related work 15 reaching consensus on the vocabulary in use. Uschold and King point out that three different strategies can be used to identify concepts and relations: bottom up (starting with most domain speciﬁc concepts), top down (starting with most general terms and moving to more speciﬁc), and middle out (starting from most relevant concepts and including both abstract and concrete terms) [86]. The bottom up allows a high level of detail in domain-speciﬁc knowledge, but may result in increased inconsistency (e.g a domainspeciﬁc concept like Heliocentrism can be grouped into several abstract concepts like Scientific Doctrine or Theory). The top down approach controls better the level of detail, however, imposing top categories early may create constraints in certain domains. Uschold and King argue that the middle out approach is the most appropriate to perform the building of the ontology because it shows a good balance between a high level of detail and a stable model. Gruninger and Uschold also afﬁrm that starting with the most important concepts ﬁrst and then defying higher level concepts leads to more stable ontologies [86]. Producing unambiguous textual deﬁnitions for each term is aimed at informally expressing the main concepts and relations which are intended to be used. Similarly to METHONTOLOGY, Uschold and King suggest the establishment of a concept dictionary. However, it should include references to the previously entered terms rather than simply dictionary-like deﬁnitions. In this way, links between terms are captured, as illustrated with the the textual description of Pseudo science: Pseudo Science is a class. It is a Field of Study which differs from Science as it can not be empirically veriﬁed. Coding involves explicitly representing the knowledge previously acquired. It consists of three main tasks: (a) committing to the vocabulary (b) choosing a formal language to express concepts and their relations and (c) programming. Uschold and King suggest a wide range of existing languages that can be used for encoding the ontology, such as Prolog, Conceptual Graphs, Ontolingua and several languages from so-called KL-ONE Family (Back, Loom, Classic) [87]. The integration of existing ontologies can be performed in parallel with the capture and coding tasks. Finally, the ontology development process ends up with the evaluation and documentation phase. The former is not speciﬁcally addressed, as stated by the authors themselves [87], however they propose to make use of competency questions for evaluating the ontology against the developed scenario. Documenting ontologies is signiﬁcant to reach a high level of knowledge sharing. Systematic guidelines for each process will turn out to be beneﬁcial in case changes are required. With regard to this, Uschold and King suggest keeping track of any set of historical notes. Chapter 2 Background and related work 16 The major drawback of Uschold and King’s method is the sudden jump from knowledge acquisition to coding without using an intermediate representation. This makes the method less applicable to large scale ontology development, where the knowledge in the ontology should be as much as possible independent from any technical aspect. Therefore, the use of intermediate representations, like Gruninger and Fox suggested, can be helpful for re-engineering complex ontologies. However, Uschold and King’s method illustrates that in small scale ontology development there may not be a need of rigorous intermediate representation. Furthermore, although Uschold and King do not provide sufﬁcient description of the techniques and activities, their suggestions, such as identifying potential users to help deﬁning the scope of the ontology, using middle out approach for identifying the main concepts in the domain, and creating dictionaries with the main concepts and relations, appear applicable to the ontology development in our case study. 2.4.4 Ordnance Survey’s methodology Ordnance Survey, the UK’s national mapping agency, has developed a methodology for the development of a conceptual ontology with the active involvement of domain experts (geographers in this case). The methodology has been followed for the development of a Hydrology Ontology. The Ordnance Survey’s methodology recognizes two main stages in ontological development-conceptual and logical-which produce two different representations of the same ontology. The conceptual ontology is deﬁned in natural language, while the logical ontology encodes the domain knowledge in a suitable formal ontology language. Mizen et al. [67] argue that this two-stage methodology can provide an efﬁcient work-ﬂow in which the domain expert creates and validates the conceptual layer and the ontologist transforms the acquired knowledge into a machine-interpretable representation. The Ordnance Survey’s methodology assigns systematic steps which domain experts need to follow to draw up the conceptual ontology [25], including: • Identifying the scope, purpose and requirement of the ontology; • Gathering source knowledge and relevant documents (books, articles, dictionaries, web resources); • Recording the captured knowledge in a Knowledge Glossary and populating it; • Converting knowledge contained in the glossary into structured English sentences; • Evaluation and documentation. Chapter 2 Background and related work 17 Similarly to Gruninger and Fox [43], the Ordnance Survey’s methodology requires identifying the scope of the intended ontology and using competency questions to check the completeness and consistency of the ontology [25]. In contrast to Uschold and King’s method [87], the Ordnance Survey does not explicitly deﬁne a range of categories of users to whom the ontology is addressed. The knowledge acquisition process already begins from the competency questions themselves due to the fact that in order to formulate them, some sort of pre-conceptualization about the core concepts of the ontology has to be expressed. Then, appropriate acquisition techniques based on informal and formal text analysis are employed in order to extract knowledge which will appear initially in the form of semi-structured sentences [67]. Those sentences should describe in natural language the purpose of the chosen term, its description and its relevant attributes or properties. The next stage is to create a glossary of terms with natural language categories for the glossary headings to assist the domain experts in ﬁlling the appropriate information. For each term, the ontologist is asked to assign a related linguistic category which can be helpful for primarily recognizing whether those terms are concepts, relations, etc. Commonly concepts are nouns (e.g. Theory in the History of Science ontology) and relationships between them are deﬁned as verbs (e.g. Investigate). In addition, it is recommended to investigate if existing ontologies have recorded identical core concepts and to compare their descriptions. If existing ontologies can be used, Kovacs et al. [25] suggest to make note of their names and locations under a column headed called Existing ontologies. The domain expert has to make a distinction between core and secondary concepts. The former are central components of the ontology and should be deﬁned appropriately, whereas the latter describe further speciﬁcation of the core concepts but do not explicitly belong to the domain and should not be deﬁned in the ontology. Secondary concepts are not fundamental for the domain of interest and sometimes can be slightly out of scope, however, they enable linking the core concepts to external domains. Table 2.3 illustrates two glossary entries from the History of Science: Term Hypothesis Synonym Conjecture Natural language deﬁnition A unveriﬁed proposal intended to explain certain facts or observations Any device employed for measuring, recording, controlling which requires skill for the proper use Linguistic term Noun Conceptual Ontology term Concept Core/Sec Core Instrument Device Noun Concept Secondary Table 2.3: Examples of Core and Secondary concepts from History of Science Ontology, deﬁned according to the Ordnance Survey’s Methodology Chapter 2 Background and related work 18 In a second dictionary, additional information about relationship terms should be recorded together with their natural languages deﬁnitions, related rules, constraints and assumptions and their occurrences in other ontologies. In case of the presence of applicable inverse (e.g. Is invented by), they should be entered only if they have been judged relevant for the domain. The next step focuses how the organization of the acquired knowledge into a set of structured English sentences will be formally encoded at the logical layer. The structured English sentences generally consist of a basic structure called relationship-term compounding by a subject concept, a verb and an object concepts (e.g. ’A model describes a phenomenon or a set of phenomena’ in which model and phenomenon are concepts). The relationship terms are reﬁned by type (inheritance, multiple inheritance, mereology, etc) and rules applied (inverse, transitive, symmetric, etc). Sentences can involve more complex structures including the use of modiﬁers (‘and’,‘or’), conjunctions (‘if’, ‘that’), adverb and adjective (‘primarily’, ‘very’). To generate structured sentences, the domain expert links each concept recorded in the ﬁrst knowledge table to a relation contained in the second table. After deﬁning structured sentences, a concept network is built in order to visualize nodes (concepts) and links (relations between them) and its graphical representation (e.g. conceptual map) depends upon the decision of the domain expert. The Ordnance Survey’s methodology adopts a two-way representation which combines network diagram and conceptual ontology triples recorded as subject-predicate-object [67]. Distinctively, Mizen et al. [67] do not encourage the use of hierarchical classiﬁcation because they believe that richer inference can be achieved paying more attention at the relation-level rather than at the taxonomical classiﬁcation [67]. The domain expert takes an active part in the evaluation, determining if the conceptual ontology is consistent and accurate to be formally encoded by the ontologist. Logical consistency, conceptual accuracy and minimal ontological commitment are the main criteria to be addressed. The consistency is determined by checking if all concept and relation terms are deﬁned in the knowledge glossary and vice versa. Semi-structured sentences must be evaluated against the structured ones. To verify accuracy, the domain expert should agree with all captured information in the ontology and, if possible, compare with the view of another expert in the same area. Our methodology for ontology development (described in Chapter 3) has been inﬂuenced by the Ordinance Survey’s methodology, which establishes a strong involvement of the domain expert in the ontology development. However, certain steps appear inapplicable in our case study, as they are tailored to having different people acting as domain Chapter 2 Background and related work 19 experts and ontologists, while these roles are performed by one person here. In particular, our case does not need a rigorous intermediate representation and structured sentences to convey knowledge from the domain expert to the ontologist. The above review identiﬁed processes which appear applicable to ontological development in this thesis. Our review aimed at ﬁnding an appropriate way of involving the domain expert throughout the entire ontology development, and particularly focusing on activities during the conceptual phase. The review pointed at commonalities and differences in the approaches, and enabled us to identify main processes to be included in our methodology which will be presented in Chapter 3. Next in the chapter, approaches for dealing with temporal speciﬁcations will be reviewed in order to provide the foundation of a framework for modeling time in a History of Science ontology. 2.5 Modeling time Modeling and representing time concepts is an issue which arises in a wide range of disciplines: artiﬁcial intelligence (AI), computer science, linguistics, philosophy, psychology, etc. The challenging aspect is the temporal contextualization of concepts and relations within ontologies in order to reason about changes upon events and actions happening over time. In this section we will review theoretical approaches for dealing with time based on key contributions from Philosophy and AI. Secondly, time modeling approaches in ontologies differing in type and shape-upper ontologies, time ontologies and history ontologies- will be reviewed. 2.5.1 Formal representations of temporal information There is a wide range of general AI theories of time which address the semantics of temporal events. From a philosophical point of view, a distinction is made between absolute and relational theories of time [60]. Absolute theories consider time according to the Newtonian notion of time [74] which does not involve anything else except its own existence. Consequently, axiomatizing time does not require further extensional knowledge with regard to the world. On the other hand, relational theories argue that temporal structures are relevant as our life is characterized by timed-related events. Thus, we need to investigate events and their properties to deeply understand the structure of time. Lin stresses the need for an intermediate theory deﬁned as Absolute Moderate [60]. According to that theory, time still remains an independent structure, while temporal primitives gain a relevant Chapter 2 Background and related work 20 importance as they need to represent our view of events as much as temporal relations have to reﬂect relationships existing between events. Although the relational approach is most appealing because it is driven by the human perception of time, both the absolute and absolute moderate approaches have also gained popularity in the AI community. The reason for this might be related to the common habit to see time as independent entity where individuals are located and events occur [92]. From a logical point of view, theories of time can be subdivided with regard to their primitives that are the basic units in any framework of reference. On the one hand, there are formalisms with explicit reference to time points and a linear ordered temporal systems. For instance, most work in Philosophy and in Situation calculus [65] and the work of Mc Dermott [66] are essentially point-based theories. On the other hand, there are models considering temporal intervals and relations between them as primitives, e.g. Allen’s interval algebra [3], Ferguson [5] and Hayes’ theories [6]. Also, several approaches for addressing the ontological nature of events have dealt with the identiﬁcation of events, and can be categorized into two groups: the unifying approach supported by Davidson [16] and Galton [33] and the multiplying approach by Goldman [36] and Kim [53]. According to Vila [92], there are three ways in which time is introduced in logic: modal temporal logics, ﬁrst order logic with temporal arguments, and reiﬁed temporal logics. The term temporal logic has been broadly used to cover all approaches to the representation of temporal information within a logical framework. This refers to the modal-logic type of approach introduced by Prior [79] under the name of Tense Logic and subsequently developed by logicians, philosophers and computer scientists. The approach is an extension of the propositional or predicate calculus with modal temporal operators with the aim of re-interpreting the notion of possible-worlds in semantics [55]. The ﬁrst order logic with temporal arguments extends functions and predicates using temporal arguments which express the time at which they need to be read. To illustrate, following [92], the sentence ‘Galileo observed sunspot on 3rd December 1611’ can be represented as follows: observe(Galileo, sunspot, 3.december.1611) The reiﬁed temporal logics allow gaining high level of expressiveness in ﬁrst order logic with regard to the truth of the assertions. This approach uses a meta-language in which formulas in the original language, the ﬁrst order logic, become terms or propositional terms in the new language. For instance, following [92] a reiﬁcation of the previous example ‘Galileo observed sunspot on 3rd December 1611’ can be deﬁned as: holds(observe, Galileo, sunspot, 3.december.1611) Chapter 2 Background and related work 21 In the next sections we will present Allen’s interval algebra and Davidson’s theory of events which are arguably the most widely used theories for dealing with time and events. The two theories will be combined in the framework for dealing with time in a History of Science ontology developed in this thesis. 2.5.2 Davidson’s theory of events Davidson’s contributions cover a wide range of topics focusing on developing a theory of meaning which will be adequate to natural language. We will not give an exhaustive account of this aspect of Davidson’s philosophy, however, we will outline it due to its relevance for Davidson’s analysis of the logic form of sentences and events. Davidson presents a theory of meaning [17] developed on the basis of a holistic conception of linguistic understanding. Davidson’s semantic theory integrates the commitment to the notion of Holism and a compositional approach according to which the meaning of sentences is dependent upon the meanings of their single parts [64]. The primary focus of Davidson, in fact, are sentences, instead of distinct words, with the purpose of building a systematic account of the core structures of the language. Davidson’s theory of meaning considers that for every potential sentence in whatever idiom, a theorem can be generated to deﬁne what those sentences mean. Since in natural language the number of possible sentences is potentially inﬁnite, we might state an inﬁnity of theorems aiming at correlating sentences to each other which are provable from a set of ﬁnite axioms [64]. Particularly, such a theory of meaning applies so-called object language to sentences in the language in which the theory of language is itself expressed with the so-called metalanguage4 . For instance, consider stating theorems that appropriately translate the English sentence ‘Galileo invented the telescope’ to an unlimited number of idioms. Following the Davidson’s view, such theory represents the framework of reference through it is possible to judge the correctness of the logical form of any discourse [58]. Therefore, a theory of meaning, requires to take the form of a truth theory of the language within sentences and events are described. Davidson’s proposal is to treat action verbs as introducing an implicit existential quantiﬁer over events, and to treat adverbs as introducing predicates of the event variable thereby included. The method proposed by Davidson is called event-token reiﬁcation, in account explicitly refers to Tarski’s theory of truth and the notion of Convention T. In fact, Tarski suggested to deﬁne the truth predicate, 4 Davidson’s by providing for every sentence s in the object language, a matching sentence p in the metalanguage that is a translation of s [64] Chapter 2 Background and related work 22 which each event-forming predicate is enriched with an extra-argument to be ﬁlled with a variable ranging over a number of particular dated occurrences (event-tokens). Relation reiﬁcation is an approach to temporal reasoning which has been fairly popular within the AI community over the past 20 years [34] [93]. It postulates that expressing a n-ary relationship with unary and binary predicates, a high level of expressiveness can be gained since relations or proprieties are treated as entities on their own. To illustrate, we will consider the sentences ‘Copernicus observed an eclipse in Rome in 1497’ and ‘Therefore, Copernicus observed an eclipse in 1497’ expressed using the event-token reiﬁcation’s method: (∃e)(Observe(Copernicus, eclipse, e) ∧ Place(e, Rome) ∧ Time(e, 1497)) ⇒ (∃e)(Observe(Copernicus, eclipse, e) ∧ Time(e, 1497)). The advantage of using the above apparatus is dealing with validity of inference within the ﬁrst order predicate logic and without making use of any external logical apparatus. Davidson’s theory of events has been widely employed in a number of AI representation frameworks based on reiﬁed events [7]. For instance, this approach has been taken into account in the Event Calculus by Kowalski and Sergot [54] and in Vila and Reichgelt’s work [93]. Davidson’s theory of events will be used in our framework for dealing with time in the History of Science for deriving a systematic approach for representing time and place speciﬁcations of History of Science events, e.g. observing a phenomenon, writing or publishing a book, inventing an invention. 2.5.3 Allen’s Interval algebra Allen describes a temporal representation that takes the notion of interval as the main primitive. His temporal theory starts with the primitive object time period, which is the time associated with some event occurring or some property holding in the world. The deﬁnition of event provided by Allen and Ferguson [5] refers to events as primarily linguistic and cognitive in nature. This implies that the world does not produce any events, but their status is determined by the human way to organize the world and deal with changes in time. Allen believes that events are arbitrary descriptions of circumstances requiring to be categorized in some way. However, he asserts that no one description is more correct than others: they may be more informative with the aim to predict analogous circumstances. Allen and Ferguson [5] distinguish between event and state. Intuitively, a state describes aspects that do not change, rather than changes over time. On the other Chapter 2 Background and related work 23 hand, from a strictly linguistic point of view, a state holds, whereas an event occurs. So, asserting that ‘John holds the door shut for a couple of minutes’, we can see that the door during this time does not undergo any change in state since it remains shut. Nevertheless, holding a door occurs and is considered a proper event even if it does not cause any modiﬁcation in terms of state. These issues have been studied extensively in semantics of natural language sentences [5], in Philosophy of language [59], and in Philosophy of mind [58]. There is a broad consensus [91] [70] in considering sentences such as ‘Kepler is German’ as descriptions of states. On the other hand, ‘Galileo observed the sunspot’ is an event involving ongoing activities. They are both true over an interval of time, however only the former holds either over intervals or subintervals. Therefore, being German is true over the interval and its subintervals. On the other hand, observing something over an interval involves some modiﬁcations from either the observer or the observed occurring over its subintervals. Events described as occurrences over intervals of time in Allen’s framework can not be reduced to a set of proprieties attached to instantaneous time points. Allen considers an interval as a collection of ordered pairs of time points. He deﬁnes thirteen basic relationships expressing the possible relations that two deﬁnite intervals can have: precedes, meets, overlaps, finished by, contains, starts, equal, started by, during, finishes, overlapped by, met by, preceded by. Each one is graphically deﬁned by a diagram relating two intervals, a and b, sorted by the degree to which a begins before b and then considering the degree to which a ends before b. Six pairs of relations are converse and whenever the ﬁrst relation is true, its converse is true also. For instance, the converse of ’a precedes b’ is ’b is preceded by a’ or ’a overlaps b’ and ’b is overlapped by a’. Allen deﬁnes the thirteen interval relationships expressed as a conjunction of point relationships [4]. A simple disjunction at an interval level, such as E1 overlaps E2, becomes much more complex expressed in a point based representation in the form of e1<e2<l1<l2 where (e1,l1) and (e2,l2) are two different event times [4]. The complication arises when uncertainty in terms of indeﬁnite intervals is introduced. Therefore, those intervals whose exact relation may be uncertain are described as sets of all basic relations that may apply, a so-called a general Allen relation. Allen’s relations are characterized by three main properties: they are distinct, exhaustive and qualitative [3]. Distinct and exhaustive because any pair of deﬁnite intervals are described by no more than one of the relations. Qualitative, rather than quantitative, because no numeric time spans are considered. Allen’s interval algebra will be used in our framework for dealing with time in the History of Science, namely for deﬁning the rules Chapter 2 Background and related work 24 apparatus for reasoning about time relations between events, (see Chapter 5). In chapters 4 and 5 we will illustrate advantages, in terms of expressiveness and applicability for historical domains, of combining both Davidson’s theory of event and Allen’s interval algebra in a temporal framework for dealing with time in a History of Science ontology. Davidson’s approach will be followed for representing time dimensions in events (see Chapter 4), while Allen’s approach will be used for deﬁning reasoning upon events by considering them as time intervals (see Chapter 5). The remaining part of this section will review how time has been modeled in ontologies, considering general upper level ontologies, time upper level ontologies, and historical ontologies. 2.5.4 Time in upper ontologies We will ﬁrst examine how and to what extent time speciﬁcations in upper ontologies can be employed in the History of Science ontology developed in this project. Upper ontologies (aka top level ontologies) attempt to describe general concepts that are the same across all domains. They usually provide a hierarchy of entities which do not belong to any particular domain and associated rules. The aim is to gain a single-shared data model which needs to be universal and articulate. Both features are intended to be relevant for correctly linking terms in existing domain ontologies to the upper ontology. We will illustrate how time is considered in upper level ontologies referring to two examples of top ontologies (Cyc and Sowa). Due to space limitation, linguistic ontologies such as Wordnet,5 SENSUS,6 or GUM,7 will not be considered here. Those ontologies provide more language-oriented approaches for deﬁning ontologies, while Cyc and Sowa illustrate how time categories are formalized in top level ontologies which is closer to our goal of formalizing the domain of History of Science. Moreover, we will also be able to consider formal approaches to conceptualizing the domain which will be useful when deciding how to represent the main categories in the History of Science. Time categories in Cyc.8 Cyc’s Upper ontology is implemented in the Cyl language, (a combination of frame and ﬁrst order logic axioms) and it is part of the Cyc knowledge base which includes more than 2.2 million common sense knowledge predicates about web page: http://wordnet.princeton.edu/ (Visited, May 2006). Concept Ontology: http://www.isi.edu/natural-language/projects/ONTOLOGIES.html (Visited, June 2006, the demo of Sensus is not available any longer). 7 Generalized Upper Model Ontology: http://www.fb10.uni-bremen.de/anglistik/langpro/webspace/jb/gum/ index.htm (Visited, June 2006). 8 The ﬁgure in this section is taken from : Condillac Knowledge Management Group, University of Savoie http://www-lgis.univ-savoie.fr/condillac/fr/ (Visited, June 2006). 6 Sensus 5 Wordnet Chapter 2 Background and related work 25 the world [57]. Concept names in Cyc are called constants. Cyc maintains a fundamental difference between individuals and collections. The former are class membership of terms expressed using isa predicates. Individuals are instances of classes, e.g. gravity and planetary motion are instances of phenomenon. The latter represent subclassof relationships in the form of genls-predicates. For instance, we might employ genls for declaring that ‘all telescopes are inventions’ and making the Galilean telescope as an instance of the collection of all telescopes. Following the argument by Matuszek et al., the distinction between individuals and collections can be useful when dealing with historical domains [22]. Although historical knowledge seems to be simply focused on individual events, places, persons, etc, philosophical and scientiﬁc knowledge is often described in terms of the properties of an entire class. For instance, saying that ‘all telescopes are inventions’ refer to the fact that they are all human-made. Cyc is formed by 43 classes and subclasses called topical groups. As shown in Figure 2.1, TimeInterval and Event are subclasses of the concept TemporalThing and Intangible. TemporalThing is a subclass of Individual, whereas Something Existing is subclass of TemporalThing which is divided in PartiallyTangible and PartiallyIntangible. Finally, IntangibleExistingThing is something PartiallyIntangible and Intangible. Figure 2.1: Temporal categories in the Cyc’s class taxonomy. Chapter 2 Background and related work 26 Time categories in Sowa’s ontology.9 Sowa’s ontology is the result of a wide range of discipline-oriented approaches coming from Logic, Linguistics, Philosophy and Artiﬁcial Intelligence [83]. Sowa proposes a lattice structure for concept type hierarchy (see Figure 2.2). A type called Universal containing all possible instances is at the top and a concept so-called Absurd type, which does not have any instances and is subclass of every concept of the taxonomy, appears at the bottom. Every pair of concepts in the taxonomy has at least a common direct or indirect superclass, as well as a direct or indirect subclass (see Figure 2.2). Hence, by combining the upper level types it is possible to represent more specialised types, e.g.: Process = Actuality ∩ Occurent Figure 2.2: Primitives in Sowa ontology. The types Process and Causality are the primitives which refer to temporal speciﬁcations. Processes can be described as any kind of change taking place between a certain start and end point. As shown in Figure 2.3, Process is subdivided by the distinction of continuous and discrete changes. The former takes into account the nature of the physical processes characterized by incremental changes over time with an explicit starting point (Initiation), Ending point (Cessation), and one whose endpoints are not being considered (Continuation). The latter considers typical computer program environments where changes occur in discrete steps called events, which are interleaved with periods of inactivity (states). The general time categories in upper ontologies appeared too abstract for the purpose to develop a History of Science Ontology which is intended to describe domain-oriented 9 The ﬁgure in this section is taken from http://www.jfsowa.com/ontology (Visited, July 2006). Chapter 2 Background and related work 27 Figure 2.3: Sowa’s Process type. knowledge. Although the level of granularity addressed in upper level was considered inappropriate for our speciﬁcation of time, it was important to consider the different use of temporal representation depending on the type and coverage of the ontology. To identify suitable time classes for our ontology, we will look at general categories described in upper time ontologies, which will be reviewed in the next section. 2.5.5 Upper ontologies for temporal concepts Time ontologies can be seen as a ‘snapshot’ of an upper ontology. They speciﬁcally deal with time concepts and relationships which can ﬁt to any domain-oriented ontology. We will consider the OWL Time ontology, developed within the DARPA Agent Markup Language project, and the KSL ontology, developed at the Knowledge Systems Laboratory at University of the Stanford. Both ontologies illustrate adaptations of temporal logic approaches based on Allen’s interval algebra. Time categories in OWL Time.10 The OWL Time ontology aims to describe temporal content of Web pages and temporal properties with regard to Web Services [49]. Its hierarchical structure includes the following temporal classes: Instant and Interval, subclasses of Temporal Entity; Proper Interval, subclass of Interval; and Instant Event and Interval Event subclasses of Event. An Instant is deﬁned as a precise point in time and an Interval is a period of time between two instants (which can be different or the same). A Proper Interval is a period of time bounded by a start and end point which are not identical. Based on the notion that an event is anything which occurs, Interval Event represents a time span between two ﬁgure in this section is taken from: http://iandavis.com/blog/2005/04/owl-time-ontology (Visited, July 2006). 10 The Chapter 2 Background and related work 28 instants, while Instant Event includes all instantaneous events. The OWL Time ontology provides the main topological temporal relations subdivided in instant-based relations (before and after) [76]. These relations can be derived for intervals too, based on the relations between their start and end points. Four predicates atTime, during, holds, timeSpan are used for linking time and events. OWL Time includes the class Eventuality to represent any event, state, process that can be located with respect to time. For instance, the predicate atTime relates an Eventuality to an Instant expressing the precise time in which it takes place. On the other hand, the predicate during speciﬁes the duration of an Eventuality in terms of each interval and subinterval it is part of. The term Eventuality is not deﬁned in the OWL Time ontology. According to Hobbs and Pan [76], another ontology should supply a clear description of the actual events. In this view, OWL Time is not in charge of the features of the Eventuality, as they will be deﬁned either in a general ontology of events or in a speciﬁc domain-ontology. The following ﬁgure illustrates the OWL time hierarchy: Figure 2.4: OWL time ontology: hierarchy Time categories in KSL.11 The Knowledge Systems Laboratory (KSL) ontology has 11 The ﬁgure in this section is taken from http://www-ksl.stanford.edu/people/ﬁkes/cs222/1998/Time/sld0 01.htm (Visited, July 2006). Chapter 2 Background and related work 29 been developed for large-scale knowledge systems application [98]. KSL is a temporal logic ontology based on the notion of continuous time line where either time points or time intervals are primitives. Consequently, the ontology class hierarchy is based on Time Point and Time Interval classes which have been made as disjoint classes. Furthermore, Time Interval subsumes ConvexTime Interval and NonConvex Timeinterval, corresponding to connected and unconnected intervals on the time line, respectively. With the purpose of handling regularly recurring events such as ‘Kepler observes Mars every Monday at 8 am’ class RegularNonConvex TimeInterval, has been included as a subclass of NonConvex Interval. For instance ‘every Monday at 8 pm’ consists of a number of connected intervals each of which representing a Monday. These connected intervals contained in one instance of NonConvexTime Interval are also single instances of ConvexTime Interval, grouped within its CalendarMonday subclass. Figure 2.5: KSL ontology’s hierarchy. From a point-based approach, TimeInterval would be the class of all sets of all points in the time line where its subclass Timepoint consists of each single point. In this line, the KSL ontology complies with the concept of Time interval in Allen’s formalization where starting point and ending point of time interval are distinct and the relations consist in disjointed standalone structures [3]. KSL also expresses the entire set of Allen relations on Time interval and 3 binary relations on Time point: (before, after, and equal-point). The KSL ontology includes further speciﬁcations on the TimeInterval into OpenInterval and ClosedInterval. The use of OpenInterval is required for describing experiments which are not composed of discrete states, for instance, to deter- Chapter 2 Background and related work 30 mine the positive velocity of a freely falling body in a Galilean experiment.12 However, if we want to assert that a given body has nonnegative (zero velocity) velocity, the concept of Open Interval would be cumbersome to deal with discrete state. The time ontologies provide a consistent and rigorous way for adding time speciﬁcations to domain ontologies. However, it can be noted that sometimes the level of granularity used in these ontologies to describe time constructs may lead to cumbersome deﬁnitions and unnecessary detail for domain ontologies. Along this line, a sub-ontology of OWL Time was developed [75] to provide most of the basic temporal concepts and relations domain applications would need. For instance, it includes a vocabulary for expressing facts about topological relations among instants, intervals, and events, together with information about durations, dates and times. Also, it allows temporal predicates to be applied directly to events. This ontology will be followed in deﬁning the time concepts in our ontology, which will be described in Chapter 4. 2.5.6 Modeling time in History ontologies Modeling time and reasoning about time dependency is one of the basic requirements to be considered when developing historical ontologies. In this section, we will review how time has been modeled in existing history ontologies, and will identify what elements of each approach can be applicable in our case study. Modeling time in VICODI. The Visual Contexualization of Digital Content (VICODI) project deals with temporal concepts combining an ontological engineering method and a fuzzy temporal reasoning model. Firstly, we will present VICODI’s ontological structure to explore how the domain is conceptualized and how time elements have been treated. We will, then, outline the VICODI temporal model pointing at its approach to dealing with imprecise information in historical domains. The examples provided in this section have been found in VICODI literature and, where appropriate, have been reformulated with regard to the History of Science domain. The VICODI ontology instance model has been developed using the Kaon framework, an extension of the W3C RDF standard.13 It consists of a superconcept, called Flavor, which has seven subconcepts: Object, Individual, Social Group, Organization, Location, Time, Event and Abstract Notion. Flavor subsumes all the main concepts in the ontology in which the instantiated relationship is time independent. VICODI considers instances related to time linked to the class Time velocity means it is moving upward, and negative velocity means it is moving downward as time increases 13 RDF/W3C Semantic Web activity: http://www.w3.org/RDF/ (Visited, June 2006) 12 Positive Chapter 2 Background and related work 31 Interval (subclass of Time). With regard to time dependent relations, VICODI applies relation reiﬁcation by coding relations as instances of Timed Relation class (subclass of Time Dependent). The use of reifying statements may sometimes cause problems. For example, it can be straightforward to assert subconcepts of persons like King or President where Henry VIII was a king and Bill Clinton was a president. However, the difﬁculty arises because no one is likely to be President or King for his entire life and in that sense representing time dependency of those relations does not seem to be an easy task. According to the notion of identity advocated by Guarino and Welty [45], VICODI introduces the Role concept including the various roles which an individual holds over time (e.g. being a student or a manager).14 Besides Person, the Role concept is applied to all time dependent instantiated relationships such as those on Objects or Abstract notions and allows attaching them to fuzzy set values. Although fuzzy and probabilistic approaches have been employed for extending description logic formalisms [28], they seem to be appropriate to deal with uncertainty and subjectivity with regard to historical facts and periods, whereas they might lead to odd and unusual results when applied to time dependent relations such as being a King or a President (for instance,‘someone can be a King with 0.8 probability’). VICODI OI model allows annotating relationships between instances incorporated at time elements level (e.g Kepler works with Brahe with annotation that this event ‘Take place at’ certain time). Their ontology model allows connecting two instances as two place predicate, e.g. (work with somebody, Kepler, Brahe). In order to avoid problems of undecidability with regard to arbitrary arity they solve this by introducing the concept of Working with somebody as subconcept of the Event concept with properties Who, Where, Whom, etc. For instance, we may attach properties, such as location or people involved to Working with concept in order to expand information on each fact. With regard to the temporal model, VICODI employs the notion of fuzzy interval to deal with the three main causes of impreciseness in history: vagueness, uncertainty and subjectivity [73]. Unlike temporal reasoning models [92], the VICODI approach deals with imprecise temporal speciﬁcation using absolute dates. With regard to this, its fuzzy temporal model takes interval as primitive where i- and i+ are respectively crisp start and end points which may not be precise. Such an imprecise interval belongs to a fuzzy set I which is deﬁned with a membership function. For example, the events which occur at a non continuous time line (e.g. the formulation of the Heliocentric theory) can notion of identity, discussed by Guarino and Welty, implies the existence of rigid and anti-rigid entity properties. For instance, the condition of being a person holds over time (rigid property), while every instance of a student might cease being a student (anti-rigid). 14 The Chapter 2 Background and related work 32 be formalized as a subset of events expressing all parts forming the master event (e.g. the Heliocentric theory was formulated for the ﬁrst time in a given time point, then was refused in several occasions, and was ﬁnally widely accepted). To achieve the same result as traditional temporal models, VICODI extends classical temporal interval relations introduced by Allen to apply to fuzzy sets. This task appeared quite challenging since VICODI’s fuzzy model does not deal with crisp interval and consequently it does not make use of crisp relations. Since there are no precise intervals, it is unlikely to determine if an interval precedes or follows another one. VICODI developers proposed a re-formulation of the notion of crisp temporal relations avoiding any reference to interval endpoints. Consequently, they introduce additional unary relations with the aim of building relations between parent and children intervals. Since complex historical notions, such as ‘Middle Ages’ or ‘French revolution’, which still cause debates in the scientiﬁc community, do not have clear start and end points, VICODI’s idea is to treat them as instances within the ontology layer and to represent a time interval as a fuzzy set of all possible time points. The membership function of a speciﬁc point represents the degree of conﬁdence that this point belongs to the fuzzy interval. Although the History of Science ontology, developed within this project, does not exploit multiple approaches for handling time categories and elements, it is important to be aware of different ways to deal with these issues. Particularly, the use of reiﬁed relations in history ontologies, as well as the extension of Allen’s interval algebra, suggested us the development a temporal framework which would combine an adaptation of Davidson’s theory of events and Allen’s thirteen relations, as discussed in chapters 4 and 5. However, the VICODI project reported limitations in terms of time modeling and use of concrete domains [68]. Concrete domains (datatypes) are widely used for representing values such as strings, numbers, integers, etc. Using the Kaon suite,15 VICODI developers experienced some problems in retrieving all instances linked to a speciﬁc instance within a speciﬁc time-frame, as it was not possible to perform this task unless checking manually the condition attached to each of them. This limitation still exists in the implementation of Kaon 2,16 the successor of Kaon framework, and some open issues still exist such as the capability of handling large number in cardinality statements and nominals [69]. Such information was considered relevant when a choice in terms of ontology language needed to be taken. Furthermore, the fuzzy approach for dealing with time in VICODI was delivered after the project ended and has not been applied for practical reasoning upon the ontology. This 15 Kaon 16 Kaon tool suite: http://kaon.semanticweb.org/ (Visited, June 2006). 2 implementation: http://kaon2.semanticweb.org/ (Visited, July 2006). Chapter 2 Background and related work 33 points out that a careful consideration with regard to reasoning upon the ontology should be taken prior to adopting the fuzzy mechanism proposed in VICODI. This, as well as the need to keep this project focused (i.e. on dealing with time and not considering vagueness, uncertainty, and subjectivity), we have decided not to include fuzzy categories in our approach. However, some of the ideas implemented in VICODI with regard to treating relations as events and considering event properties, appear applicable to modeling time in a History of Science ontology and will be followed in our work (see Chapter 4). Modeling time in SWHI. Semantic Web for History (SWHI) is an ongoing project dealing with history ontology development for assisting general users, not only historians, in exploring the American history by using the enhancements of Semantic Web technology. SWHI’s ontology has been built using VICODI at its core and extended with Newsbank Topic Hierarchy (NTH), the topic classiﬁcation scheme in SWHI knowledge base, FOAF17 an RDF vocabulary describing people and the links between them, and selected concepts from metadata schemas (DC, MARC21). VICODI was used as out of box ontology to model time-dependency and temporal speciﬁcations. For example, all instances of persons are grouped under foaf:Person with added properties by using vicodi:hasRole. Scientist, Natural Philosopher or Astronomer are deﬁned as properties of foaf:Person. Time speciﬁcations are expressed using the vicodi:exists property which links instances (i.e Galileo Galilei) to existing time instance (1564-1642). The value 1564-1642 is attached to vicodi:Time and its subconcepts making use of the property vicodi:exists [97]. The notion of time in SWHI follows two different views: linear and concentric. From a philosophical perspective, linear time is a continuous ﬂow moving horizontally from past to present and future. It portrays time as an absolute physical reality and asserts that the passage of it is independent of consciousness. In contrast, the notion of concentric time seems to be related to a Leibnizian tradition asserting that we inscribe events and situations within certain abstract apparatus [56]. Concentric time is expressed in discrete units, such as years, centuries and decades, while linear time is based on conceptual units, such as era, movements, historical periods and events. For example, linear time is represented by including year units in the metadata ﬁelds [97]: <datafield tag="651" ind1="" ind2="0"> <subfield code="a"> Connecticut </subfield> <subfield code="x"> Politics and government </subfield> <subfield code="y"> 1775-1865 </subfield> </datafield> The following example illustrates metadata representing concentric time encoding the 17 Friend of Friend Project home page: http://www.foaf-project.org/ (Visited, May 2006). Chapter 2 Background and related work 34 fact that Queen Anne’s War spans in the period of 1702-1713. This concentric data represents an event encoded as an instance of vicodi:Event: <datafield tag="651" ind1="" ind2="0"> <subfield code="a"> United States </subfield> <subfield code="x"> History </subfield> <subfield code="y"> Queen Annes War, 1702-1713 </subfield> <subfield code="v"> Personal narratives </subfield> </datafield> Although SWHI’s work does not speciﬁcally aim at modeling time, it deals with history ontologies and their requirements identiﬁed in Section 2.2. The notions of linear and concentric time was considered relevant and inﬂuenced the earliest stages of our time representation, as discussed in Section 4.3. In contrast to SWHI and VICODI, we will employ ﬁrst order logic in our formalism for representing and reasoning about time dependencies in the History of Science. Modeling time in HICO and TELOS. The History of the Iranian Constitution ontology (HICO) aims at expressing the semantic content of historical documents with the purpose of posing and resolving historical queries [88]. In line with historical domain requirements, time modeling is one of the main issues which HICO deals with. HICO aims at capturing temporal relations between concepts, as well as modeling hierarchy changes over time. For instance, the ‘governmental position hierarchy’, expressing roles in the Iranian institution, in HICO will be considered as an entire structure to which changes may happen any time considering changes in the roles taking part. With regard to this, HICO refers to temporal relations expressed in TELOS [71]. TELOS is a knowledge representation language which provides a model for capturing time evolution and semantics in applications. It grounds its representation on the notion of time interval by considering time relation and their inverses for expressing potential relationships between intervals. The temporal relations are from Allen’s interval algebra, however the TELOS framework applies a few modiﬁcations to these, including (a) conventional dates and time (e.g. 2006/10/05, indicating the 5th of October 2006), (b) semiinﬁnite intervals with conventional dates and times as one endpoint (e.g. 2006/05/12* stands for one of the boundaries of the an imprecise interval) and special intervals such as Alltime or Now allowing a language based restriction of temporal assertions [71]. The example below, taken from [71], shows how TELOS embeds temporal knowledge in metadata descriptions of information resources: TELL TOKEN martian IN Paper (at 1986/10..*) WITH: author firstAuthor: Stanley (at 1986/10..*); : LaSalle (at 1987/1..*); Chapter 2 Background and related work 35 : Wong title (before 1987/5) : ’The MARTIAN system’ END The IN clause in the above example establishes martian to be an instance of Paper for an unbounded interval time which start in October 1986. Then WITH asserts that Stanley has been the author value over the duration expressed by 1986/10..*. Then the start interval when the following person happened to be an author during the interval boundary 1987/1..*. TELL, UNTELL and RETELL clauses are introduced for expressing belief times in the system, which are associated in TELOS to every proposition in the knowledge base and are distinct from the actual state of the world. For instance, a given fact might have been changed in January, but the knowledge base will maintain the previous information as long as it is not upgraded [71]. The TELOS framework does not express a set of relationships between intervals, it only enables one to represent the single temporal relationship attached to each of the temporal components of a given deﬁnition. The HICO ontology gives the idea of how challenging it is to deal with historical resources and their attached time-related concepts, showing emerging open issues in which Semantic web technologies can take action. HICO illustrates an application of the representation formalism TELOS which validates some of the main ideas in that framework. Relevant to our case study is the TELOS approach for dealing with intervals following Allen’s interval algebra. However, the representation used appears rather speciﬁc and tailored to adding metadata to historical resources. It was considered as inappropriate in our case study. 2.6 Summary In this chapter, we outlined the major challenges involved in modeling historical domains which guided us in determining the nature and characteristics of our case study: a History of Science ontology. We considered modeling time to be the main challenge addressed in this work. In order to identify a suitable methodology for building a History of Science ontology, we have reviewed a number of existing methodologies: METHONTOLOGY, Gruninger and Fox’ methodology, Uschold and King’s method and Ordnance Survey’s methodology. Although none was considered to be fully applicable to our domain, we were able to identify appropriate elements and aspects which were re-used during our ontology development. Chapter 2 Background and related work 36 In Section 2.5, approaches and frameworks for dealing with temporal information were presented. Among those, we further discussed Davidson’s theory of events and Allen’s interval algebra. Their integration was considered to be appropriate for representing and reasoning on the History of Science domain, as discussed in Chapter 4 and 5. Davidson’s account of events served as the basis for systematically dealing with time inclusion and Allen’s thirteen relations were used to compare events occurring at different time. While Allen’s approach has a wide range of application in the A.I. and computer science communities, Davidson’s theories have not witnessed a similar impact on those communities [7]. A substantial part of Section 2.5 illustrated how temporal relations are represented in ontology-based applications. We considered three categories of ontologies differing in the subject of the conceptualization and in the structure, as advocated by Van Heijst and colleagues [89]: upper ontologies (Sowa, Cyc ), time ontologies (OWL TIME and KSL) and history ontologies (VICODI, SWHI, HICO and TELOS). While Sowa and Cyc’s contribution of this review was aimed at completeness, time and history ontologies were relevant ontology-based exempliﬁcation of temporal approaches. However, even time and history ontologies were not directly applicable in our work. For instance, although VICODI makes use of reiﬁcation in order to add temporal properties to relations, it does not provide a uniﬁed and systematic approach to identify, embed and reason on time reﬁnements. In Chapter 4 we will describe our approach for modeling time in the History of Science ontology developed in this thesis. The main time elements and predicates belonging to this ontology will be outlined, together with the temporal model in use (Davidson’s event token reiﬁcation). In Chapter 5, querying and reasoning on the domain will be illustrated through the combination of the Davidson’s theory of event for representing temporal reﬁnements and Allen’s interval algebra for reasoning on temporal relations. Chapter 3 Building a History of Science Ontology: Main Phases 3.1 Introduction In this chapter, we will present the methodology we have followed for building a History of Science ontology. It combines aspects from the methodologies reviewed in Chapter 2. A walk through the identiﬁcation and encoding of the main categories and relations will be provided to illustrate the main activities and tasks. The resultant ontology1 is given in Appendix A, and full glossary entries are provided in Appendix B. The task for creating an ontology in this project is different from existing approaches for creating ontologies. Not only is the nature of the domain speciﬁc, as discussed in Section 2.3, but also the fact that the author was the domain expert who also acted as ontologist had to be taken into account. The implications of this were twofold: on the one hand, we did not have to deal with communication issues across different expertises, on the other hand, we needed to make sure that the domain expert developed appropriate knowledge to translate the conceptual ontology into a formal language and to compose competency questions. Based on the fact that both roles, expert and ontologist, were attributed to the same person, it was not possible to appropriately separate the roles during the ontology construction process. Roughly, the activities requiring mostly knowledge entire ontology is, also, available through the following URL: http://www.comp.leeds.ac.uk/ilaria/hiso.pl 1 The 37 Chapter 3 Building a History of Science Ontology: Main Phases 38 about the History of Science can be considered as part of the role of a domain expert, while the coding activities can be attributed to the ontologist role. In the following sections we will present a three-phase methodology including preconceptualization, conceptualization, and logical phase. These phases correspond to the distinction between conceptual and logical ontology made by the Ordnance Survey methodology reviewed in Section 2.4.4. The pre-conceptualization phase (described in Section 3.2) aims at clarifying the ontology scope and its requirements. The conceptualization phase concerns the identifying of main concepts and relations and is usually done informally. This phase is outlined in Section 3.3 describing the main classes and relations in the ontology. The logical phase, which is outlined in Section 3.4, is the formal modeling of the domain, and includes ontology encoding, veriﬁcation, and tuning. More details for the last two phases will be given in the following chapters. The description of conceptualization will be extended in Chapter 4 where we will present an approach for modeling time aspect in History of Science, while Chapter 5 will provide detail about querying the ontology used for ontology veriﬁcation and tuning during the logical phase. 3.2 Pre - conceptualization As shown in the methodologies discussed in Section 2.4, the clariﬁcation of the ontology’s requirements and scope is a prerequisite for the identiﬁcation of main categories and relations. In order to identify the ontology scope and to prepare the conceptualization phase, we conducted three main activities which composed the pre-conceptualization phase: • Providing a high level description of the domain and its features; • Building a scenario; • Identifying a range of potential informal competency questions to be addressed. The pre-conceptualization phase, in fact, overlapped with the background review process for this thesis. We conducted research to identify the nature of History of Science and the main characteristics of this domain. The result of this research was presented in Section 2.3. We chose to describe high level characteristics common across historical domains. Although the description of the domain was not directly connected to ontology creation, the main characteristics of the domain helped us clarify what aspects would be most important in our case. In particular, the decision to focus on modeling time and Chapter 3 Building a History of Science Ontology: Main Phases 39 reasoning about relations between events and people, and to consider dealing with subjectivity and vagueness outside the scope of our study, was critical for deciding what concepts to include and how to encode domain relations. This allowed us to narrow down the features we wanted in order to address them when we came to specify motivating scenarios and informal competency questions. It is important to point out that at this stage, we started collecting a range of references covering the following purposes: • Eliciting the main characteristics of the domain; • Investigating existing projects involving the use of ontologies in historical domains. We made use of references at different levels of detail throughout the pre-conceptualization and conceptual phase [2] [11] [15] [61] which gave us a general understanding of the features to be accomplished and acted as background reading for building our scenario. Following the argument in [43], we have built a motivating scenario with the purpose of outlining a set of challenges and potential ontology-based solutions. An initial deﬁnition of our scenario was set in parallel with the overview of the domain. However, we reﬁned it several times, even during the main phases, to check our conceptualization against the scenario problems. The reﬁned version of the general illustrative scenario, which was used for motivating our work and for identifying the scope of the ontology, was given in Chapter 1 where we discussed the use of an ontology-based representation to structure the access to digital resources. One of the most signiﬁcant steps in deﬁning the scope of the ontology, which also led to reﬁning the scenario, was when we narrowed down the domain of interest. Because History of Science is a fairly broad domain, to keep this work focused and to accommodate the time constraints, we decided to select a sub-domain related to a particular topic of investigation: the Scientiﬁc Revolution in Europe between 16th and 19th centuries and the Astronomical Revolution. Following Uschold and King’s method (see Section 2.4.3), within the scenario, we deﬁned a range of intended users (audience) to whom we addressed our solutions. Initially, we wanted the ontology to cover different needs of a heterogeneous group of users. However, time constraints required us to focus on one speciﬁc category. The background reading played an important role to suggest the proﬁciency and skills our users were required to have. We looked at philosophical and historical resources for both inexperienced [85] [46] and advanced users [12] [20] to proﬁle the targeted users we need to consider. We discarded the idea of considering researches or lectures in History or Phi- Chapter 3 Building a History of Science Ontology: Main Phases 40 losophy of Science2 as it would have required too high a level of specialization of the ontology. Also, it was difﬁcult and imprecise to identify skills and competencies valid for the whole professional community. We then looked at educational domains following informal discussions with educational experts who pointed at recent reports about decline in the students’ interests in science. A way to entice students to science is to help them discover facts and dependencies between scientiﬁc events3 . In that respect, knowing about the History of Science is important. Hence, an ontology-based approach that facilitates learning about the History of Science and searching through corresponding digital resources was considered as beneﬁcial. The main purpose of our ontology is to provide the foundation for such an approach. So, we looked at some educational resources, and speciﬁcally we consulted the A-level syllabus in Philosophy [9] in order to understand to what extent the History of Science was presented. This content matched the level of details we were aimed to achieve, and thus, we recognized the A2 level students as intended users for our ontology. The last step in this phase was to formulate a provisional list of informal competency questions to address within the scenario with regard to the intended users and application of the ontology. According to Gruninger and Fox’s methodology (discussed in Section 2.4.2), we used competency questions for clarifying the scope of the ontology and, more importantly, when performing the conceptual and logical phases. We will give examples of competency questions and will discuss their application for verifying and tuning the ontology in Chapter 5. On the whole, the pre-conceptualization phase was entirely characterized by the role of a domain expert. We managed to move from a broader domain description to a narrow sub-domain covering a speciﬁc topic. This prepared the phase of identifying of the main concepts and relations. Although the author acted as the domain expert, the preconceptualization phase was more time consuming that initially planned because further reading was required to scope the ontology by considering the balance between temporal constraints and the appropriate level of details. Furthermore, the amount of work involved for envisioning the scenario and the intended users was signiﬁcant. However, a clear understanding of the main characteristics of the domain and the role of the competency questions enabled us to better scaffold the remaining phases. The next section will describe the conceptualization phase which was rooted in the scenario and the features of the domain. The activities and tasks undertaken were driven do not discuss about the difference between the two disciplines, however they have many notions in common for which it was important to consider both of them during our background reading 3 Nest research report, Real Science - Encouraging experimentation and investigation in school science learning: www.nesta.org.uk/assets/pdf/real science report NESTA.pdf (Visited, May 2006) 2 We Chapter 3 Building a History of Science Ontology: Main Phases 41 by the choice to speciﬁcally address time and dependencies between events and people. 3.3 Conceptualization The conceptualization phase included two major activities: • Activity 1: Identifying main concepts and relations; • Activity 2: Identifying time concepts and including time dimensions in relations. Activity 1 will be presented in detail in this chapter, while Activity 2 will be discussed in the next chapter where we present our approach for dealing with time in the History of Science Ontology. Identifying the main concepts and relations in a History of Science domain included several tasks which were often undertaken in parallel: • Investigating existing ontologies for reusing terms and relations; • Acquiring knowledge from different sources; • Drafting a concept tree and relations in the form of triples; • Building separated tables for concepts and relations. On the whole, these tasks were not executed in a sequential order. Most of the time, several tasks had to be performed in parallel as they were to some extent complementary. For instance, investigating existing ontologies was considered to be integrated with acquiring knowledge from other sources because both tasks aimed to locate appropriate concepts and relations. The former is aimed at partially reusing available ontologies appropriate for our purpose. We will present them in the next section together with a number of examples of our formalization in Prolog. The latter involved consulting literature in Philosophy and History of Science, such as [47] [48] [84] [1] which provided the main vocabulary of the domain. Also, we made use of a History of Science dictionaries [10], on line English dictionaries and thesaurus 4 5 , a free-content encyclopedia such as Wikipedia6 and a lexical databases like Wordnet 7 to check the appropriate use of certain terms in the domain. For instance, we needed to check the the difference in meaning 4 Merriam 5 Wiktionary: Webster available at http://www.m-w.com/ (Visited, June 2006). http://en.wiktionary.org/wiki (Visited, May 2006). 6 Wikipedia portal: http://www.wikipedia.org/ (Visited, April 2006). 7 Wordnet:http://wordnet.princeton.edu/ (Visited, June 2006). Chapter 3 Building a History of Science Ontology: Main Phases 42 between Theory and Law or the use of the terms Method and Model in Science. The descriptions of these concepts are given in Appendix B. After acquiring the basic knowledge, we provided a draft of a concept tree in which main categories were simply grouped as classes and subclasses. Also, we included the most signiﬁcant relationships expressed as predicates with two arguments. For example, a relation between two concepts - Theory and Phenomenon - was represented as: explain(theory, phenomenon). It should be noted that some formalization was introduced in the conceptualization phase. This was possible because the domain expert acted as an ontologist as well. Because of this, we were able to introduce ﬁrst order predicate logic notations without considering any communication issues between the two roles. We ﬁrst attempted to identify main concepts and relations simply writing down them in a text document. Once the skeleton of the taxonomy and the main relations were identiﬁed, we validated them against a set of competency questions, as deﬁned in Section 3.2 . We built three separated tables for concepts, relations and instances. Similarly to METHONTOLOGY and Ordnance Survey methodology, we organized the concept table based on the following headings: Name of the concept, Type, Synonyms (if appropriate), Natural language description, Source (if appropriate). Table 3.1 illustrates three concepts within our ontology: Name Observation Type Concept Synonym – Natural Language Description act of recognizing and noting a fact or occurrence often involving measurement with instruments the act or process of discovering the act or process of inventing Source Merriam Webster/Wordnet/SUMO Merriam Webster/Wordnet/SUMO Science ontology/Merriam Webster/Wordnet Discovery Invention Concept Concept – – Table 3.1: History of Science Glossary of Terms: Example Entries Similarly to METHONTOLOGY, our relations table includes binary relations in which a source concept is linked to a target concept. However, we did not assert the mathematical properties (transitive, inverse, or symmetrical relations) which were added at the logical phase, as explained in Chapter 5. Table 3.2 shows three entries in our relation table holding respectively the concepts Experiment - Theory, Person - Document and Person - Person: The instances table contains all instances attached to a certain concept. Table 3.3 shows some instances of the concept astronomer. Chapter 3 Building a History of Science Ontology: Main Phases Relation name Challenge Publish Inﬂuence Source concept Experiment Person Person Target concept Theory Document Person 43 Table 3.2: History of Science Relations Table: Example Entries Concept: Astronomer Instance (1): Ptolemy Instance (2): Kepler Instance (3): Copernicus Table 3.3: History of Science Instances Table: Example Entries The last activity of the conceptualization phase was to build time inclusion and event tables. The former is aimed at identifying relations in which time elements had to be embedded, whereas the latter visualizes our approach to modeling time, based on Davidson’s theory of events, as discussed in detail in Chapter 4. We considered natural language definitions describing the purpose of relations involving time, natural language examples in which time may occur and informal queries which later served for querying the ontology. To encode event relations, we built an event-relations table. Table 3.4 illustrates a ternary relation holding three additional properties in terms of begin, end and location of a given event: Event name Investigate – – Source concept Galileo – – Target concept sunspot – – Event identiﬁer d galileo investigate sunspot – – Event property begin: 1612-04-00 end: 1636 location: Italy Table 3.4: History of Science Event Relations Table: Example entries In this section we presented the main activities of the conceptualization phase. We can consider this phase as the one which required a balanced involvement either of the domain expert or the ontologist. The deﬁnitive identiﬁcation of the main concepts and relations have been a long-term process in which the hierarchy achieved a stable structure only in the latest stages. In the next section, we will present the way we encoded and represented those concepts and relations according to the formalism in use: Prolog. 3.4 Logical representation and coding The logical phase was subdivided in four main activities: • Choosing the ontology language; Chapter 3 Building a History of Science Ontology: Main Phases 44 • Converting glossary knowledge into formal representation; • Querying and reasoning about the domain. In order to chose the appropriate language for our domain, we looked at a range of existing languages. According to Corcho et al. [14], there is an increasing number of languages which underly the development of the Web: so called web-based ontology languages or ontology markup languages. With the advent of XML as standard language for interchanging information on the Net, a number of derivate languages were built on top of it. SHOE8 was made XML-syntax, XOL9 enables one to build concept taxonomies and binary relations, RDF and RDF Schema also include some constraint checking reasoning. Later, OIL, DAML+OIL and OWL were developed as semantic extensions to RDF and RDF Schema, adding Description logic primitives to RDF. We initially considered the use of OWL as the language to encode our ontology. However, following the experience of the most outstanding project in history ontology, VICODI [72] (see Section 2.5.6) which pointed out that RDF based formalisms appeared less expressive for modeling historical domains, we decided to look at other formalisms. Frame-based paradigm or the combination of frame and ﬁrst order logic (Ontolingua, OCM, Flogic) focus primarily on representing classes (called frames) and their properties (called slot), rather than relations between them. Moreover, the author was unfamiliar with existing frame-based formalisms. Hence this family of languages were considered not fully appropriate for our purpose. Finally, we looked at logic-based approaches for building ontologies in terms of logic programming languages like Prolog and its implementations (Quintus, SICStus, SWIProlog). Prolog was found suitable as the language for encoding the ontology because we could separate the data model (describing how data was represented) from the data description (actual facts from the ontology). The presence of this abstract level was signiﬁcant especially during the querying and reasoning process with regard to: • Using recursive querying techniques; • Applying depth-ﬁrst and breadth-ﬁrst searches strategies. Converting our glossary of concepts, relations and instances into Prolog syntax was the next step. The concepts populating our glossary table were encoded in Prolog as classes and subclasses. According to Deswarte and Oosthoek [19], we attempted to limit the number of upper classes with the purpose of avoiding a unmanageable number of 8 SHOE 9 XOL Project: http://www.cs.umd.edu/projects/plus/SHOE/ (Visited, June 2006). speciﬁcation:http://www.ai.sri.com/pkarp/xol/ (Visited, June 2006). Chapter 3 Building a History of Science Ontology: Main Phases 45 relations. However, we could not meet the requirements of no more than seven upper classes as suggested by the authors. The main reason was that dealing with different levels of granularity (domain, subdomain and topic concepts) justiﬁed to broaden the number of main classes. In order to illustrate our hierarchy and the way we encoded the identiﬁed concepts in Prolog, we ﬁrstly present the upper classes providing the reasons for their inclusions and if appropriate the corresponding sources. Secondly, we examine the subclass path of one of the most populated class we encoded. The entire class and subclass hierarchy is given in Appendix A, together with all glossary entries (Appendix B). The ontology is also available through the following URL: http://www.comp.leeds.ac.uk/ilaria/hiso.pl Both classes and subclasses were encoded as Prolog facts. The notation introduced consists of atomic formulas containing respectively one and two constants, as follows: class(person). subclass(scientist, person). The following concepts were considered to be the upper classes in our ontology. class(person). class(role). class(place). class(phenomenon). class(’field of study’). class(belief). class(document). class(’mode of reasoning’). class(doctrine). class(event). class(method). class(model). class(’group of people’). class(time). class(d_e). Person, Role, Place and Document can be seen as general classes aiming at categorizing the most representative protagonists of the our sub-domain (the Scientiﬁc revolution), their roles and the places where lived and have undertaken their scientiﬁc activities. Similarly, Group of people class was encoded to deal with sub-set of people having some commonality. For instance, we considered Academic organization, Non academic organization, Philosophical school and Scientific school to be direct subclasses of Group of people. subclass(’academic organizaton’, ’group of people’). subclass(’non academic organization’, ’group of people’). subclass(’philosophical school’,’group of people’). subclass(’scientific school’,’group of people’). Chapter 3 Building a History of Science Ontology: Main Phases 46 Using the subclass predicate, we deﬁned further levels of classes, e.g.: subclass(university, ’academic organization’). In Chapter 5, we will deﬁne rules for inferring subclasses that have not been encoded explicitly in the ontology. For instance, we can derive that university is a subclass of ’group of people’. As mentioned earlier, existing ontologies were partially reused in few cases during the ontology development. For example, with regard to Role, Group of people and their subclasses, we reused concepts taken from Role,10 VICODI11 and Relationship ontology.12 Furthermore, we encoded classes which could be considered to be domain-speciﬁc, such as Phenomenon, Event, Model, Method and Mode of reasoning. These classes pertain to the History of Science as discipline dealing with approaches for the observation and explanation of real world phenomena. Science ontology 13 and its subentry ontology in astronomy were employed to organize classes such as Event, Method, Model, and in general those related to the scientiﬁc inquiry. In addition, Doctrine class has been inﬂuenced by Michele Pasin’s ontology for philosophical resources 14 . We include this class and its path with the purpose of highlighting that the origins of the Science are rooted in philosophical questions and investigations. However, we attempted to make a distinction between philosophical and scientiﬁc doctrine: subclass(’scientific doctrine’, doctrine). subclass(’philosophical doctrine’,doctrine). subclass(realism, ’philosophical doctrine’). subclass(rationalism,’philosophical doctrine’). subclass(idealism, ’philosophical doctrine’). subclass(pragmatism, ’philosophical doctrine’). subclass(phenomenology, ’philosophical doctrine’). subclass(dualism, ’philosophical doctrine’). subclass(monism, ’philosophical doctrine’). subclass(analytic, ’philosophical doctrine’). subclass(intuitionism, ’philosophical doctrine’). subclass(materialism, ’philosophical doctrine’). subclass(mecanicism, ’philosophical doctrine’). subclass(mentalism, ’philosophical doctrine’). ontology: http://www.hypergrove.com/OWL/Role/index.html (Visited, June 2006). ontology: http://herakles.fzi.de/vicodi/vicodi2 0040402.html (Visited, April 2006). 12 Relationship ontology: http://vocab.org/relationship/ (Visited, July 2006). 13 Science ontology: http://archive.astro.umd.edu/ont/index.html (Visited, June 2006). 14 Michele Pasin ontology: http://kmi.open.ac.uk/people/mikele/philontology/owl/philosophy.owl, (Visited, July 2006). 11 VICODI 10 Role Chapter 3 Building a History of Science Ontology: Main Phases 47 The evident limitation is related to the fact that we were not able to identify scientiﬁc doctrine per se. On the contrary, we encoded a relevant number of subclasses of philosophical doctrine. We will discuss the need of populating the ontology in Chapter 6 together with the involvement of a second domain expert for revising and validating the appropriateness of the ontology structure. Class Time is discussed in detail in Chapter 4. It has been deﬁned as a general class for describing temporal concepts. The vast majority of time concepts belong to OWL time ontology,15 KSL and AKT support ontology,16 as we will illustrate in Chapter 4. The class d e is related to modeling additional speciﬁcation in the form of events. The need to include this class and its usage will be discussed in detail in Chapter 4 which will present our approach to dealing with time. Finally, Belief aimed at describing notions which were considered to be truth without any scientiﬁc evidence. An example of belief is the notion of ‘crystalline spheres’, that is the concentric and eccentric revolving spherical transparent shells in which the stars, sun, planets, and moon were supposed to be set.17 The class Fields of study relates to the various areas of study, such as: subclass(humanities,’field of study’). subclass(science,’field of study’). subclass(’pseudo science’, ’field of study’). We also included the class Pseudo science to consider ﬁelds of study like Astro logy which do not have the status of scientiﬁc disciplines. Because during the Scientiﬁc revolution many scientists were conducted studies in Astrology, for consistency, we considered this subclass to be part of Field of study. To encode instances, we used a similar notation but encoded the atomic formulas with the (fact instance of) predicate. It links instance name and the corresponding class to which the instance belongs, e.g.: fact_instance_of(sillogism, ’deductive argument’). fact_instance_of(’crystalline sphere’, hypothesis). It can be noted that Crystalline sphere is both an instance of belief and hypothesis. A hypothesis, in fact, is a suggested explanation of a phenomenon which needs to be validated. time ontology web page: http://www.isi.edu/ pan/OWL-Time.html (Visited, July 2006). support ontology (Ontolingua environment): http://d3e.open.ac.uk/akt/2002/support-ontoling-v2.0/support-ontoling-v2.0-t.html (Visited, August 2006). 17 This deﬁnition is taken from: http://dict.die.net (Visited, December 2006) 16 AKT 15 OWL Chapter 3 Building a History of Science Ontology: Main Phases 48 Following our notation, the relations (data model) and instantiated relations (data) were recorded by using predicates relation type and fact relation, respectively. The relation type predicate deﬁned the name of the relation and the class of its source and target concepts, as illustrate in Table 3.2. Concepts holding a relations can be also the same. The following examples show relation type and their corresponding instantiated relations: relation_type(extend, theory, theory). fact_relation(extend, ’galilean theory’, ’kepler theory’). relation_type(write, person, document, d_e). fact_relation((write, ’Galileo’, ’Dialogue Concerning the Two Chief World systems’, d_galileo_write_dialogue). It can be noted that the second relations contain four constants instead of three. This will be justiﬁed and explained in detail in Chapter 4, where we show how to deal with temporal properties by considering relations with Davidson’s event as a parameter. The last activity in the logical phase is to query and reason on the acquired knowledge. For this, we formulated appropriate rules. This activity will be presented in Chapter 5, together with a discussion on the use of competency questions for verifying the consistency and expressiveness of our ontology. Acting as a domain expert, the author, developed new competencies to efﬁciently program with Prolog and specify rules which will need to be validated against the original ontology scope and scenario problems. With regard to this, we experienced some sort of delays within the logical phase, as the author acted as ontologist, but was initially not familiar with the mechanism for coding the ontology. 3.5 Summary In this chapter, we presented our methodology for building the History of Science Ontology consisting of three main phases: pre-conceptualization, conceptualization, and logical phase. Our methodology should be considered as a domain application of the most signiﬁcant features of the methodologies, reviewed in Section 2.4. Furthermore, it must be stressed that one of the challenging features of our methodology is the broad spectrum involvement of the domain expert which includes the coding activity too. However, managing such different range of activities linked to the same person, it caused difﬁculties and schedule rearrangements especially when we were required to establish our reasoning framework. Chapter 3 Building a History of Science Ontology: Main Phases 49 This chapter presented in detail the pre-conceptualization phase and gave an overview of the conceptualization and logical phases presenting the main concepts and relations in the ontology. The following chapters will present how time speciﬁcations were added to the ontology and what rules were deﬁned to reason about the domain. Chapter 4 Modeling Time in a History of Science Ontology 4.1 Introduction As discussed in Chapter 2, modeling time and reasoning about temporal relationships is a key challenge in historical domains. In this chapter, we will present an approach for modeling time in a History of Science ontology. Following our methodology, outlined in Chapter 3, this section will introduce the main time classes and will suggest a way for representing time in conceptual relations. We have identiﬁed three main dimensions in which time can be expressed: time concepts, temporal relations and instantiated relations in the form of events. The ﬁrst consists of top classes declaring main temporal speciﬁcation. The second and third dimensions utilize Davidson’s theory of events for representing time respectively at the data model level (temporal relations) and at the data level (instantiated relations). Using Davidson’s theory has given us a systematic way to deal with temporal dimensions of historical relations. To illustrate the advantage of this approach, we will ﬁrst present our initial approach which was modiﬁed to take into account type-token distinction for identifying the relations in which time is included. Section 4.2 introduces the main time concepts. Our initial approach for modeling time in History of Science relations will be presented in Section 4.3, while Section 4.4 will present a more systematic way for adding time which overcomes the problems faced by the initial approach and has 50 Chapter 4 Modeling Time in a History of Science Ontology 51 been followed in the History of Science ontology in this thesis. 4.2 Time concepts in History of Science ontology This section will discuss how time has been expressed at the concept level. In line with the approaches for modeling time in ontologies discussed in Chapter 2, we consider all time categories belonging to a class Time, the superclass of all time-related classes. Time point, Time interval, and Time unit, which are subclasses of Time, represent the core time-related classes in our ontology: subclass(’time point’, time). subclass(’time interval’, time). subclass(’time unit’, time). Following a widely accepted deﬁnition, a time point is an extensionless point on the universal timeline. The time point at which a particular event occurs can be known at different level of granularity and precision. For example, the AKT ontology (see Chapter 2) deﬁnes a superclass called Time position whose aim is to describe Time interval and Time point on the time axis (referring to a time zone standard). On the other hand, the KSL ontology speciﬁes a function on the domain of Time point (Location-Of) which has range Time-Quantity that expresses the amount of time from ‘point zero’ on the timeline to the time point of interest [98]. In our ontology, instances of Time point are expressed as numbers consisting of year-month-day, (e.g. 1611-03-12) is a time point instance indicating when Scheiner observed sunspot and 1473-02-19 is a time point instance that represents the date when Copernicus was born. If the month or day are unknown, these are represented with 00. For instance, 1612-04-00 is a time point instance associated with the investigation of sunspot by Galileo which happened in April 1612, while to represent the fact that Brahe observed the supernova in 1572 (month and day unknown) we use the time point instance 1572-00-00. This special meaning of 00 is taken account of in the implementation of time comparison operations and in the temporal inference rules. We refer to the notion of Time interval as a set of time points, which is in line with the KSL ontology [98] (see Chapter 2). Although KSL speciﬁes also Convex and Non Convex intervals, as well as Open and Closed intervals, we did not consider them as part our taxonomy as they imply a much higher level of granularity than required in our case. At the initial stages we considered the use of Non Convex and Open/Closed intervals for representing respectively regularly recurrent events and Chapter 4 Modeling Time in a History of Science Ontology 52 value changes (e.g. velocity) over an interval and its subintervals. We followed the argument in [98] that Open and Closed intervals may be expressed making use of the relation meets where the domain does not require such level of granularity at the concept level. Due to time constraints and because the facts we were encoding did not require Open and Closed intervals, we discarded these categories from our ontology. It should be noted that these categories may have to be added when more complex facts from the History of Science are encoded, as pointed out in Chapter 6. The last temporal category in our ontology is Time unit which refers to the units of measurement that may be used to quantify time periods. An experiment may employ a time unit referring to seconds or minutes, rather than months or years, pointing out the importance of considering the granularity of units of measurement. A structured way to deal with this issue is illustrated in KSL (see Section 2.5.5) where the class Time-Granularity is used to specify granularity at time points (e.g Month Granularity). For instance, time point 1 January 2000 with day granularity is a single time point which can be any point within a convex time interval, starting midnight of December 31st and ending midnight of January 1st. Instead, the AKT support ontology deﬁnes Time-Measure, subclass of Unit-of-measure, as the class of all units of measures used to measure time, e.g minute, second, hour.1 In our ontology, we did not consider a rigorous way of dealing with time units because we could simply link events to a certain point in time expressed in day, month, or year. The conversion between different time units may be possibly handled at the rule level. In order to keep this work focused, we did not consider ﬁner granularity of the Time Unit. While the above time categories can be related to any ontology, we needed a temporal category that could relate to time periods in History domains, i.e. Historical period. This category was subjected to a number of revisions in terms of hierarchy and labels, especially during the earlier phases of ontology development. In order to build a consistent class hierarchy, we looked at the notion of periodization as an attempt of categorizing historical periods into discrete named blocks.2 The deﬁnition of periodization in historical subjects serves as a starting point for making appropriate decision with regard to the widely accepted time period labels. The nature of the domain itself requires some sort of temporal classiﬁcation with the purpose of making sense of the past and articulating changes over time. However, all periodization systems are to some extent arbitrary and their usage is often based on cultural connotations which are local to certain countries. For instance, Romantic Period refers to European and European-inﬂuenced support web page: http://kmi.open.ac.uk/projects/akt/ref-onto/ (Visited, June 2006). deﬁnition is from the Wikipedia entry for Periodization:http://en.wikipedia.org/wiki/Periodization (Visited, January 2007). 2 The 1 Akt Chapter 4 Modeling Time in a History of Science Ontology 53 cultures. We decided to follow the approach presented by Wilcox [96] where time periods were deﬁned by their cultural usage. According to Wilcox, historical periods are considered as cultural movements and eras (e.g. Romantic Period, Scientific revolution, Middle ages), 3 rather than particular time periods deﬁned with decimal numbering system(e.g. 17th Century). In order to identify the most notable historical periods, several sources have been consulted. Firstly, we looked at the sub-level categories in Google directory with regard to History. Google uses only Ancient and Middle ages as cultural time speciﬁcations together with a number of decimal numbering system entries. We needed a ﬁner granularity for cultural-based historical periods. Thus, we consulted sources speciﬁcally related to digital libraries with historical content. The Internet Public Library, 4 subject-categorized directory of authoritative websites, provides the functionality to browse its resources following a categorization by an era. Similarly to Google, it shows only the cultural categories Prehistoric, Ancient History and Medieval History (Middle ages), together with a number of categories ordered by discrete units, e.g. 15th Century History or 16th Century History. Finally, we consulted the Echo web site, 5 a resource center for History of Science, Technology and Industry, enables one to search or browse through its content not only by topic or keyword, but also by time period. The the most recent version of our taxonomy shown below follows the core time periods in Echo. Following the argument in [19] that modeling a high number of classes within an ontology leads to an unmanageable number of relations, we decided to limit historical period classes to those encompassing the most general and signiﬁcant ones. We consider the following subclasses of the class Historical period: subclass(’historical period’, time). subclass(ancient, ’historical period’). subclass(’middle ages’, ’historical period’). subclass(’early modern’, ’historical period’). subclass(modern, ’historical period’). Speciﬁc historical periods are represented as instances of the main classes. For example, the Middle ages has three further speciﬁcations deﬁned as instances: 3 We may also name periods based on inﬂuential individual who played a relevant role, e.g. Victorian Era. 4 Internet Public Library web site: http://www.ipl.org/ (Visited, January 2007). 5 Echo Portal: http://echo.gmu.edu/center.php (Visited, March 2007). Chapter 4 Modeling Time in a History of Science Ontology 54 fact_instance_of(’early middle ages’, ’middle ages’). fact_instance_of(’high middle ages’, ’middle ages’). fact_instance_of(’late middle ages’, ’middle ages’). Examples of instances of the classes Early Modern and Modern include: fact_instance_of(’scientific revolution’, ’early modern’). fact_instance_of(’industrial revolution’, modern). The process of identifying classes and their instances was driven by the use of informal concept deﬁnitions [52]. As discussed in our methodology (Chapter 3), natural language descriptions assist the development and the population of the ontology. In order to reach a well-established Historical period hierarchy and to generate a number of related instances, we consulted free-content encyclopedias like Wikipedia,6 general purpose dictionaries such Cambridge or Merriam-Webster7 and domain oriented dictionaries [47] [21]. The Wikipedia entry below presents an informal concept deﬁnition of the Early Modern period. It can be noted that the function of this textual description is basically to assist the developer to build a consistent ontology skeleton in which wider historical periods are narrowed down into a number of signiﬁcant instances. The early modern period is a term used by historians to refer to the period in Western Europe and its ﬁrst colonies which spans the time between the Middle Ages and the Industrial Revolution that has created modern society.[...] The beginning of the early modern period is not clear-cut, but is generally accepted to be in the late 15th century or early 16th century.[...] The end date of the early modern period is usually associated with the Industrial Revolution, which began in Britain in about 1750. In this section, we introduced the temporal categories encoded in our ontology and pointed at similarities and differences with time ontologies reviewed in Chapter 2, Section 2.5.5. The next step in modeling time is presenting time dimensions in relations across the History of Science domain. Our initial approach to adding time into conceptual relations is outlined in the next section. 4.3 Modeling time in relations: initial approach In order to add time to domain relations (second dimension), we tried to identify the recurrent time-patterns occurring at the relation level by collecting examples illustrating portal: http://www.wikipedia.org/ (Visited, February 2007). primarily consulted the on line version of Merriam Webster available at http://www.m-w.com/ (Visited, February 2007) 7 We 6 Wikipedia Chapter 4 Modeling Time in a History of Science Ontology 55 when time had to be included. Based on the examples, we identiﬁed two main categories of time appearance: ‘durativity’ and ‘punctuality’. Following Comrie [13] and Frawley [32], we considered durativity to refer to events which can be predicated over time, whereas punctuality characterizes events which have no temporal duration and are not temporally distributed.8 According to Comrie [13], we, also, introduced the notions of ‘repeatable’ and ‘non-repeatable’ events which relate to both durative or punctual forms of time.9 The term ‘repeatable’ refers to events which are repeated, whereas ‘non-repeatable’ is associated to events happening only once (e.g. being born, died, inventing an invention).10 With regard to this, we noticed that the majority of non-repeatable forms of time across our domain would appear in two different cases: as relational predicate (‘was born at’) or as instantiated relations (invent, Galileo, telescope). The description below illustrates an example used to identify a particular need of non-repeatable time inclusion: We wish to know when a certain person holding an occupational role was born or died. Assuming we have encoded the fact that Galileo is an astronomer, we can derive that Galileo is an instance of person. Then, the non-repeatable time relations across the domain requires corresponding relations for when somebody was born and died referring to an instance of time point. According to Vendler’s classiﬁcation [90], the lexical aspect of verbs, so-called Aktionsart,11 plays a crucial role in the way they relate to time. Verbs are said to be divided into two main categories: atelic and telic verbs. While the former refers to verbs that do not have natural endpoints (e.g. being born does not involve any goal in its semantic structure), the latter considers verbs expressing actions or events as tending toward a particular accomplishment (e.g. extending a theory over a certain period of time). Correspondingly, in our initial approach to modeling time in domain relations, we added parameters of class Time point to all non-repeatable relations occurring at a unique time. The example below shows how we encoded the time when a person was particular, Comrie distinguished between imperfectivity and durativity. The former relates to the internal structure of a situation or event in terms of duration, phasal sequence, etc. While, the latter simply considers any events which last in time. The notion of durativity was considered appropriate for our needs 9 For instance, Comrie [13] considered the example of ‘cough’as punctual verb which can be either referred to a situation taking place only one (one single cough) or to a situation which is repeated (a series of coughs). 10 In Comrie [13], repetable/non-repeatable distinction corresponds to semelfactive and iterative forms of time, respectively 11 As advocated by Binnick [8] and Comrie [13], Aktionsart is commonly regarded as a category by which lexical items can be sorted in terms of internal temporal constituency. 8 In Chapter 4 Modeling Time in a History of Science Ontology 56 born, died, or invented an invention: relation_type(’was born at’, person, ’time point’). relation_type(’died at’, person, ’time point’). relation_type(invent, person, invention, ’time point’). By contrast many relations are associated with repeatable events (e.g. a theory extends another theory, a model describes a phenomenon). An example pointing at the need to represent this is the following: A theory can extend the meaning of a previous one by means of further speciﬁcations or more accurate measurements. For instance, the Copernicus theory extends the Brahe theory of planetary motion or the Copernicus theory replaces the Ptolemy theory over a certain interval of time. It appeared sensible to add parameters of class Time interval to domain relations when repeatable time occurred, as shown in the example below: relation_type(extend, theory, theory, ’time interval’). relation_type(explain, model, phenomenon, ’time interval’). However, when trying to populate the domain facts, we realized that specifying time intervals for durative relations was inconvenient. For some of these relations, we found time dimensions in the domain sources used (e.g. when an experiment conﬁrmed a theory), while for others these appeared much more difﬁcult to be univocally identiﬁed (e.g. we could not ﬁnd time-related facts to specify when a theory extends another theory). It was noted that repeatable/non-repeatable distinction did not lead us to a systematic approach to adding time across the entire domain, and a better way for deciding where and how to add time was needed. Furthermore, adding time as a parameter to relations led to cumbersome cases at times. For example, to encode not just when but also where a person was born we had to consider two different relations associated with the event when the person was born: relation_type(’was born at’, person, ’time point’). relation_type(’was born in’, person, place). Similarly, we noted that 3-argument relations required adding (sometimes) not just time but also place. For instance, we wanted to say that a person wrote a book at a certain time point and at a certain location, or when two people worked together we wanted to specify both the time and location. It became apparent that adding more and more variants of naming for the same relation in order to handle different event dimensions was not an appropriate way. We needed to ﬁnd a systematic way for adding time dimension to relations, which had to include: Chapter 4 Modeling Time in a History of Science Ontology 57 • well-deﬁned criterion for deciding which relations would require additional temporal speciﬁcation and which would not; • well-speciﬁed mechanisms for adding time dimensions to domain relations. The next section will describe a more systematic approach for adding time to historical relations which was followed for building the ﬁnal version of our History of Science ontology. We will model relations as Davidson events (described in Chapter 2) and, instead of using repeatable/non-repeatable distinction, will consider type/token distinction in order to decide when a Davidson event has to be associated with a domain relation. 4.4 Modeling time in relations: use of Davidson events In natural language, relations are often expressed as verbs, because they are associated with actions. Following Davidson’s theory (Section 2.5.2), we decided to associate events to domain relations, as opposed to our initial approach which associated only time parameters to relations. For this, an upper class d e (Davidson event) was deﬁned to describe events associated to relations using properties, such as location, time, and duration. By using instances of Davidson events, we can add time-place related properties to instantiated relations in an uniﬁed way. In order to decide when a relation would have a Davidson event associated to it, we consider whether the relation involves types or tokens. The distinction between type and token originates in Philosophy [77] [80] [81] [95] and has been widely applicable in Humanities, as well as in scientiﬁc ﬁelds. It is generally accepted that types are abstract entities corresponding to some form or pattern, whereas tokens are usually deﬁned as particular physical manifestations of types. By applying the type/token distinction to the parameters of every relation, we could break down our relational predicates into three groups: • parameters are only types (we call these type-type relations); • parameters include a token and a type (we call these token-type relations); • parameters are only tokens (we call these token-token relations). Examples of the three types of relations include: Type-Type: Chapter 4 Modeling Time in a History of Science Ontology 58 relation_type(relate, model, ’field of study’). relation_type(extend, theory, theory). relation_type(explain, model, phenomenon). All parameters in the example relations above are universal concepts. Token-Type: relation_type(invent, person, invention). relation_type(challenge, experiment, theory). relation_type(observe, person, phenomenon). Person and experiment are tokens but invention, theory and phenomenon are types. Token-Token: relation_type(’work with somebody’, person, person). relation_type(influence, person, person). relation_type(’was born’, person). It is important to notice that some concepts can be either tokens or types. For instance, in the relation type (invent, person, invention), Invention is either a type (representing any telescope) or a token (referring to a speciﬁc invention, e.g. the particular telescope Galileo invented). A concept can be either type or a token on the basis of its domainoriented application. For instance, in our ontology we consider Invention to be type, while another ontology may distinguish between concrete inventions and may treat invention as a token. Since types are abstract patterns they do not have a physical or temporal existences, thus a relation holding between two types, it can be considered as timeless relation. Hence, it does not require an event associated with it and it is not required to add any further time-place speciﬁcations. For example, a theory extends another theory at any time and place: relation_type(extend, theory, theory). fact_relation(extend, ’galilean theory’, ’copernican theory’). fact_relation(extend, ’copernican theory’, ’brahe theory’). fact_relation(extend, ’brahe theory’, ’ptolemy theory’). The integration of type-token distinction and Davidson’s theory of events seem to reﬂect the notion of entity orders, advocated by Lyons [62] [63]. According to the author, we can distinguish three types of entities. First order entities are concrete entities which can be individuated, quantiﬁed, enumerated [63] Chapter 4 Modeling Time in a History of Science Ontology 59 Whereas, second order entities are: events, processes, states-of-affairs or situations which can be located in time and occur or take place rather than exist; e.g. continue, occur, apply12 . [94] Instead, the third order entities are ideas, thoughts, theories, hypotheses, that exist outside space and time and which are unobservable [94]. Consequently, types and token correspond to the third and ﬁrst order entities in Lyons, respectively, whereas the second order entities comply with the deﬁnition of event as adopted in this work.13 Every time a relation includes at least one token, we can potentially add temporal speciﬁcation and additional properties (e.g. location), i.e. we will associate a Davidson event to it. We will illustrate the adding of Davidson events with several examples. The ﬁrst example shows how the relation type (invent, person, invention) has been treated as a Davidson event. Following the Davidson approach, we consider invent as a three-place predicate reifying an event: relation_type(invent, person, invention, d_e). We can then add temporal or location speciﬁcations when instantiating the relations. Hence, we can encode that Galileo invented the thermometer in 1593, while he invented the telescope in June 1609 but continued its improvement until the fall of 1609. The extract from the ontology encoding this is given below: fact_relation(invent, ’Galileo’, ’thermometer’, d_galileo_invent_thermometer). event_property(begin, d_galileo_invent_thermometer, 1593-00-00). event_property(end, d_galileo_invent_thermometer, 1593-00-00). fact_relation(invent, ’Galileo’, telescope, d_galileo_invent_telescope). event_property(begin, d_galileo_invent_telescope, 1609-07-00). event_property(end, d_galileo_invent_telescope, 1609-11-00). mentioned in section 4.3. the lexical aspect of a verb identiﬁes its relation to time (e.g. telic and atelic verbs) 13 Moreover, Lyons introduces the distinction between form and expression which explicitly refers to the notion of words as tokens and words as types advocated by Pierce. For instance, the sentence ‘He who laughs last laughs longest’ [63] is composed of six word-forms and ﬁve word-expressions. Word-forms represent each instance of a given type (laughs as token occurs twice), whereas word expressions can be seen as the headwords of dictionary entries [63] (the two occurrences of ‘laughs’ are forms of the same expression that is ‘to laugh’). 12 As Chapter 4 Modeling Time in a History of Science Ontology 60 It can be noted that even if starting and ending point of an event are the same we encode both, which appeared convenient for deﬁning predicates to identify event relations, see Chapter 5. The second example illustrates how the adding of Davidson event enabled us to include time and place speciﬁcations in a consistent manner. For instance, we can encode the fact that Kepler and Brahe worked together in Prague in the period 1600 - 1601 as: relation_type(’work with somebody’, person, person, d_e). fact_relation(’work with somebody’, ’Kepler’, ’Brahe’, d_kepler_brahe_collaboration). event_property(begin, d_kepler_brahe_collaboration, 1600-00-00). event_property(end, d_kepler_brahe_collaboration, 1601-00-00). event_property(location, d_kepler_brahe_collaboration, ’Prague’). Similarly, we modiﬁed the initial deﬁnition of the relation type (‘was born at’, person, ‘time point’) (see Section 4.3) to reify a Davidson event. This enabled us to specify both the location and the time when somebody was born, e.g. we could encode that ‘Copernicus was born in 1472-02-19 in ‘Torun’ as: relation_type(’was born’, person, d_e). fact_relation(’was born’, ’Copernicus’, d_born_copernicus). event_property(begin, d_born_copernicus, 1473-02-19 ). event_property(end, d_born_copernicus, 1473-02-19). event_property(location, d_born_copernicus, ’Torun’). The third example illustrates an advantage of using the Davidson’s approach: the ability to represent temporal inclusion by considering a set of atomic events as part of a primary event. For instance, Copernicus died on 1543-05-25 and during his lifetime lived in Frombork between 1510-00-00 and 1520-00-00. The predicate sub event has been speciﬁed in order to represent temporal subsumption of events. fact_relation(died, ’Copernicus’, d_died_copernicus). event_property(begin, d_died_copernicus,1543-05-25 ). event_property(end, d_died_copernicus, 1543-05-25). event_property(location, d_died_copernicus, ’Warmia’). fact_relation(live, ’Copernicus’, d_copernicus_live_frombork). event_property(begin,d_copernicus_live_frombork, 1510-00-00). event_property(end, d_copernicus_live_frombork, 1520-00-00). event_property(location, d_copernicus_live_frombork, ’Frombork’). event_relation( sub_event, d_copernicus_live_frombork, d_born_copernicus ). Moreover, we considered events which are conceptually correlated (e.g. a given observation leads to the associated investigation of the relative phenomenon), though they do not postulate temporal inclusion as necessary condition. According to Sandstr¨ m [82], o a relation holding between two events is deﬁned as ‘response’ when the main event, e1 Chapter 4 Modeling Time in a History of Science Ontology 61 leads, provokes, inspires e2 [35] With regard to this, we introduced the predicate inﬂuential event. For instance, Galileo investigated the sunspot between 1612-04-00 and 1636-00-00 in Italy. Such investigation has been driven by observation of the sunspot which happened on 1611-03-12 in Rome. We encoded this event correlation as follows14 : fact_relation(investigate, ’Galileo’, sunspot, d_galileo_investigate_sunspot). event_property(begin, d_galileo_investigate_sunspot, 1612-04-00). event_property(end, d_galileo_investigate_sunspot, 1636-00-00). event_property(location, d_galileo_investigate_sunspot, ’Italy’). fact_relation(observe, ’Galileo’, sunspot, d_6obs). event_property(begin, d_galileo_observe_sunspot, 1611-03-12). event_property(end, d_galileo_observe_sunspot, 1611-03-12). event_property(location, d_galileo_observe_sunspot, ’Rome’). event_relation(influential_event, d_galileo_observe_sunspot, d_galileo_investigate_sunspot). Finally, we will illustrate how the adding of Davidson events enables us to describe historical periods which are treated as events. In line with the above way of representing temporal constructs, we added a relation hold to indicate that a historical period was associated with a particular Davidson event, which can then be described with appropriate properties. For instance, the Scientific revolution and the Industrial revolution have been deﬁned as follows: relation_type(hold, ’historical period’, d_e). fact_relation(hold, ’scientific revolution’, d_e_scientific_revolution). event_property(begin, d_e_scientific_revolution, 1543-00-00). event_property(end, d_e_scientific_revolution, 1750-00-00). fact_relation(hold, ’industrial revolution’, d_e_industrial_revolution). event_property(begin, d_e_industrial_revolution, 1780-00-00). event_property(end, d_e_industrial_revolution, 1850-00-00). In this section, we discussed an adaptation of Davidson’s theory of event to add time speciﬁcations to relations following the distinction between type and token concepts. In the next chapter, we will illustrate how Davidson events can be used in rules that extract knowledge from the History of Science ontology. is important to note that this is an illustrative example of how the correlation between events can be represented, and the full theory was out of the scope of this work. 14 It Chapter 4 Modeling Time in a History of Science Ontology 62 4.5 Summary This chapter has discussed an approach for representing temporal speciﬁcations in the History of Science. Our approach to time is based on a preliminary phase in which we have reviewed a number of ontologies (see Chapter 2) that differ in terms of shape and purposes. We walked through the development of our ontology presenting how time dimensions have been added, i.e. ﬁrst including corresponding time categories and then adding time speciﬁcations to relations. We presented our initial attempt of including temporal speciﬁcation to domain relations and pointed out its limitations leading to the development of a more systematic temporal framework grounded on the Davidson’s theory of events and distinguishing type and token parameters in relations. The framework presented here provides a simple, and yet expressive way for modeling time in a History of Science ontology, which can be applied to any historical domain. The uniﬁed approach for representing relations enables us to identify temporal relationships between domain events, which will be illustrated in the next chapter. Furthermore, we can write fairly general inference rule by means of Prolog predicates which are applicable to a wide range of domain relations. The next chapter will describe rules for querying the ontology and reasoning about the History of Science. Chapter 5 Querying the History of Science Ontology 5.1 Introduction Querying and reasoning about the domain enables retrieving information that matches certain criteria from the ontology. This is one of the main activities during the logical phase of our methodology (Chapter 3), and is closely linked to ontology veriﬁcation and tuning. In this chapter, we will present rules for reasoning about the History of Science, and will discuss how these rules have helped to verify and tune our ontology. We will outline the query modes across our domain, grouping them into three categories: Concept-based, Relation-based, and Time-Event-based. Each mode will be described providing a number of meta-rules for manipulating and inferring the encoded information. In order to exemplify main query modes, we will present a set of questions for writing domain-oriented rules. The questions will be categorized based on the type of outputs they produce. They can be used as competency questions to verifying the consistency and check the expressiveness of the ontology, as discussed in Chapter 3. The query presented in this chapter will illustrate extracting facts that have already been encoded in the History of Science ontology, as well as inferring knowledge not explicitly encoded in the ontology. In line with the coding formalism chosen for the ontology, we have used Prolog for deﬁning rules to query the ontology. 63 Chapter 5 Querying the History of Science Ontology 64 This chapter will ﬁrst deﬁne rules for the main modes of query and will illustrate them with example queries over the History of Science ontology (Section 5.2). Then, in Section 5.3, we will present domain-related questions which utilize the query modes. Examples of who, what, where and when and combined questions will be presented. In Appendix A, we will provide additional display predicate examples, while the entire ontology and the rules can be seen via the following URLs: http://www.comp.leeds.ac.uk/ilaria/hiso.pl and http://www.comp.leeds.ac.uk/ilaria/rules.pl Finally, section 5.4 will discuss how the queries have been used for verifying the ontology by summarizing the main problems discovered and corrections made. 5.2 Ontology-based query modes Query modes deﬁne the basic categories of queries used to interrogate our ontology. We have identiﬁed three query modes considering the inference rules employed: Conceptbased, Relation-based, and Time-event related. The Concept Mode involves rules concerning classes and instances from the domain ontology, while Relation Mode includes rules that infer knowledge by considering transitivity, symmetry and inversion of relations. The Time-Event Mode is based on the previous modes and includes reasoning about temporal and event-based speciﬁcations. We will describe each mode by presenting a number of rules applicable to the domain representation described in chapters 3 and 4. Each rule will be deﬁned as a Prolog predicate and illustrated with example queries over the History of Science ontology. We will use corresponding display predicates to illustrate the results of employing the rules to query the domain. The display predicates will use the built-in Prolog predicate setof which collects a list with the results of a query and removes the duplicates. The resultant list of answers will then printed on the screen. 5.2.1 Concept mode The Concept mode allows one to determine direct and indirect subclass relationships among classes, as well as direct and indirect class-instance relationships. We will deﬁne Prolog predicates that check for these relationships. Direct subclasses of a class are those which have been directly encoded in the ontology via the subclass predicate. For a given class, the subclasses can be derived with the following query: ?- subclass(X,class) Chapter 5 Querying the History of Science Ontology 65 This can be used to see part of the direct class hierarchy. For example, ?- subclass(X,scientist). inventor; discoverer; astronomer; mathematician; naturalist; chemist; physician; anatomist; physicist In a similar way, we can check for the direct superclasses of class by using subclass(class,X), e.g. the superclass of scientist is person. The whole hierarchy of direct classes can be retrieved by considering the subclass path predicate: subclass_path(X,[],Y):-subclass(X,Y). Going through the class hierarchy, we can infer all subclasses of a class, which enables us to check multiple inheritance links. The following predicate deﬁnes a rule for deriving inferred subclasses by considering recursively all direct subclasses and their subclasses. inferred_subclass( X, Y ) :- subclass(X, Y). inferred_subclass( X, Z ) :- subclass(X, Y), inferred_subclass(Y,Z). This predicate allows going upwards and downwards through the hierarchy. It will be used in some questions deﬁned in Section 5.3. We have also used this predicate to verify the hierarchical class structure of the ontology. For this, the following display predicates were deﬁned (showlist has been deﬁned to simply display each item and to list on the screen): show_inferred_subclasses :setof( inferred_subclass(X,Y), inferred_subclass(X,Y), All ), showlist( All ). show_inferred_subclasses_of(Y) :setof( X, inferred_subclass(X,Y), All ), showlist( All ). The result of show inferred subclasses, which is illustrated in Appendix A, is a list with all class inheritances, while show inferred subclasses of can be used to check the subclasses of a given class. For example, ?- show_inferred_subclasses_of(person). anatomist; astrologer; astronomer; chemist; cosmologist; discoverer; epistemologist; inventor; logician; mathematician; methaphysician; natural philosopher; naturalist; philosopher; physician; physicist; scientist More examples of this query are given in Appendix A. In a similar way, we can derive indirect subclasses of a given class, i.e. subclasses that are inferred but are not directly encoded in the ontology, by using the rule: Chapter 5 Querying the History of Science Ontology 66 indirect_subclass( X, Y ) :inferred_subclass( X, Y ), \+subclass(X,Y). Following the above example, the indirect subclasses of person will not include scientist and philosopher. Similarly to inferred subclasses, we have deﬁned corresponding display methods that were used for verifying the type hierarchy. Examples are given in Appendix A. Furthermore, the subclass path predicate has been deﬁned to enable the display of all direct and indirect subclasses stored in the ontology. A corresponding display predicate is deﬁned to show the subclass path. subclass_path( X, [Y | Rest], Z ) :subclass( X, Y ), subclass_path( Y, Rest, Z ). show_extended_subclass_paths :setof( subclass_path(X,[H|T],Y), subclass_path(X,[H|T], Y), All ), showlist( All ). We can use show extended subclass paths to visualize the skeleton of the ontology and to check the hierarchy path. The result of this predicate is given in Appendix A, and an extract from it is shown below to illustrate the class path of a given class. subclass_path(metaphysical cosmologist,[cosmologist],philosopher). subclass_path(metaphysical cosmologist,[cosmologist,philosopher],person). This shows that metaphysical cosmologist inherits from philosopher via cosmologist, and from person via cosmologist and philosopher. Direct instances attached to a given class can be derived by using the fact instance of predicate. For example, ?- fact_instance_of(X,astronomer). ’Hipparcus’; ’Ptolemy’; ’Kepler’; ’Copernicus’; ’Brahe’; ’Galileo’; ’Scheiner’; ’Newton’; ’Halley’; ’Huygens’ In a similar way, we can check for the class of an instance by using ?- fact_instance_of(instance,X). For example, the classes of ’Kepler’ are astronomer and matemathician. To display the whole instance structure of the ontology, we deﬁned the following display predicates: show_direct_instances:setof(fact_instance_of(X,Y), fact_instance_of(X,Y), All), showlist(All). show_direct_instances_of(Y):setof(X, fact_instance_of(X,Y), All), showlist(All). Chapter 5 Querying the History of Science Ontology 67 The result of show direct instances enabled us to check all instance inheritances, while show direct instances of can be used to check the direct instances of a given class. For example, ?- show_direct_instances_of(’natural philosopher’). [’Archimedes’, ’Bacon’, ’Democritus’, ’Euclid’, ’Gilbert’] Going through the instance structure, we can also infer all instances of a given class which enables us to check multiple inheritance chains. The following predicate deﬁnes a rule for deriving inferred instances by considering recursively all direct instances and their inferred subclasses: inferred_instance( X, C ) :- fact_instance_of(X,C). inferred_instance( X, C ) :- fact_instance_of(X,C1), inferred_subclass(C1, C). We have deﬁned corresponding display predicates to show the instance structure of the ontology. The predicate show inferred instances checks all instance inheritances, see the result in Appendix A. The predicate show inferred instances of was used to verify all inferred instances of a given class. For example: ?-show_inferred_instances_of(person). [’Archimedes’, ’Aristotle’, ’Bacon’, ’Brahe’, ’Copernicus’, ’Darwin’,’De amark’, ’Democritus’, ’Descartes’, ’Einstein’, ’Euclid’, ’Galenus’,’Galileo’, ’Gilbert’, ’Halley’, ’Harvey’, ’Hipparcus’, ’Hobbes’, ’Huygens’, ’Kepler’, ’Leibniz’, ’Newton’, ’Pascal’, ’Ptolemy’, ’Roger Bacon’, ’Scheiner’, ’Vesalius’] Note that the ontology does not explicitly specify these instances as person. In line with indirect classes, we can derive indirect instances of a given class, i.e. inferred instances that are not directly encoded in the ontology. The following predicate can be used: indirect_instance(X,C):inferred_instance(X,C), \+fact_instance_of(X,C). For example, using this predicate with a ﬁrst parameter ’Kepler’ will return scientist and person, while in the ontology ’Kepler’ is speciﬁed only as an astronomer and mathematician. Corresponding display predicates have been used to show the results of applying indirect instances, see Appendix A. Chapter 5 Querying the History of Science Ontology 68 5.2.2 Relation mode The Relation mode includes rules that retrieve the transitive, symmetrical and inverse relationships closure. To indicate whether a relation is transitive, symmetrical, or inverse, we used the following indicators: transitive_relation(relation). symmetrical_relation(relation). inverse_relation(relation, inv_relation). It is important to note that whereas the indicators for transitive and symmetrical relations are monadic, the inverse relations is a two-argument predicate (inverse relation/2). Corresponding indicators were associated with every relation in the ontology, e.g.: relation_type(extend, theory, theory). transitive_relation(extend). inverse_relation(extend,’is extended by’). relation_type(influence, person, person, d_e). inverse_relation(influence,’is influenced by’). transitive_relation(influence). relation_type(meet, person, person, d_e). symmetrical_relation(meet). relation_type(observe, person, phenomenon, d_e). inverse_relation(observe,’is observed by’). relation_type(’agree with’, person, ’group of people’,d_e). Note that a relation can be associated with any subset of the above indicators, including none (as in the case of ’agree with’. Using the indicators, we can derive relationships that have not been directly encoded in the ontology. As the above example shows, some relations have 3 and other 4 parameters. In order to unify the access to relation instances, we will use the predicate get relation: get_relation(R,X,Y):fact_relation(R,X,Y). get_relation(R,X,Y):fact_relation(R,X,Y,_). The following predicate deﬁnes a rule for deriving inferred transitive relations by considering recursively all relations which are directly encoded as transitive and their inferred transitive relations. inferred_transitive_relation( R, X, Y) :- transitive_relation(R), get_relation( R, X, Y ). Chapter 5 Querying the History of Science Ontology 69 inferred_transitive_relation( R, X, Y ) :transitive_relation( R ), get_relation( R, X, Y1 ), inferred_transitive_relation( R, Y1, Y ). Using the above predicate, we can derive inferred transitive relations for a given relation. For example: ? inferred_transitive_relation(influence, X,Y). X=’Hipparcus’,Y=’Ptolemy’; X=’Kepler’,Y=’Brahe’; X=’Copernicus’,Y=’Kepler’; X=’Copernicus’,Y=’Galileo’; X=’Copernicus’,Y=’Brahe’ The ﬁrst four inﬂuences have been directly encoded in the ontology, while the last one was derived by following the transitivity of the relation and indicates that ’Copernicus’ indirectly inﬂuenced ’Brahe’. In a similar way, we can apply inferred transitive relation to check for inclusion of places: ? inferred_transitive_relation(contain,’Europe’,Y). ’Prague’; ’Denmark’; ’Knudstrop’ The fact that ’Knudstrop’ is in ’Europe’ is inferred following the transitivity of the relation contain. The scarcity of results from this query shows the need to check for missing relation instances. For example, by adding: fact_relation(contain,’Europe’,’Italy). fact_relation(contain, ’Europe’, ’Germany’). we will be able to infer that ’Graz, ’Roma’, and ’Pisa’ are in ’Europe’. To display all transitive relations encoded in the ontology, we used the display predicate show inferred transitive, see Appendix A. In a similar way, we can deﬁne predicates to infer relationships between concepts following symmetry and inversion of relations. Symmetrical and inverse relations do not require recursive predicates because they only apply to direct relation instantiations. The following predicate can be used to infer symmetrical relations: inferred_symmetrical_relation(R,X,Y):symmetrical_relation(R), get_relation(R,X,Y). inferred_symmetrical_relation(R,X,Y):symmetrical_relation(R), get_relation(R,Y,X). Chapter 5 Querying the History of Science Ontology 70 We can use it in domain-related questions to check/derive relationship between concepts. For example, we can make queries where the order of the parameters can be swapped, like in: ?- inferred_symmetrical_relation(’work with somebody’,’Kepler’,Y). ’Brahe’; ’Wallenstein’ This takes into account that ’Kepler’ worked as an assistant of ’Brahe’ and at another time ’Wallenstein’ was assisted by Kepler. Inverse relations enable deriving knowledge by changing the relation direction from child of to parent of. The following predicate has been deﬁned to infer inverse relation by taking into account that names can be different and arguments have to swap: inferred_inverse_relation(R,X,Y):inverse_relation(R,_), get_relation(R,X,Y). inferred_inverse_relation(R,X,Y):inverse_relation(R,R1), get_relation(R1,Y,X). inferred_inverse_relation(R1,Y,X):inverse_relation(_,R1), get_relation(R1,Y,X). Ontologies generally include many inverse relations, and this is the case with our ontology. Hence, this predicate can be very helpful to check for a particular relation despite the naming used, e.g. the example below collects all relation instances of write and ’is written by’: ?-inferred_inverse_relation(write,’Galileo’,X). ’Sidereus Nuncius’; ’Dialogue Concerning the Two Chief World Systems’; ’Discourse on the Two Sciences’; ’The Assayer’ The result is the same as using the query: ?-inferred_inverse_relation(’is written by’,X,’Galileo’). Examples of using relation-based rules in domain questions are given in Section 5.3. 5.2.3 Time-Event mode The Time-Event mode includes rules that enable inferring information about events and reasoning about temporal speciﬁcations considering time points and events. To get the properties and all detail about an event, the following rule can be used. The built in Prolog predicate findall composes a list (P) with all information about an Chapter 5 Querying the History of Science Ontology 71 event. The predicate event properties is used to collect the time and location of an event. The additional information about the event (e.g. the relation and the concepts it links) are collected using the relation instantiations that refer to this event. event_properties_detail(E,P):findall(Happening, ( event_properties( E, P_e ), ( fact_relation( R, A, B, E ), Happening = [R, A, B, P_e] ) ; ( fact_relation( R, A, E ), Happening = [R, A, P_e] ) ), P). For example, the following query can be used to get more information about the event d galileo write dialogue1 : ?-event_properties_detail(d_galileo_write_dialogue,P). [write, ’Galileo’, ’Dialogue Concerning the Two Chief World systems’, [1624-00-00, 1632-00-00, ’Italy’]] To compare events, we will follow Allen’s interval algebra (see Chapter 2) which derives relations between events based on their start and end points. We have deﬁned arithmetical operators for comparing timepoints by considering the yyyy-mm-dd format followed in representing instances of the class ’time point’ (see Chapter 4). To check whether a time point is before another time point, we use the following rule: timepoint_before(Y1-_-_, Y2-_-_):-Y1<Y2,!. timepoint_before(Y-M1-_, Y-M2-_):-M1<M2,!. timepoint_before(Y-M-D1,Y-M-D2):-D1<D2,!. For example, the following queries return positive results: ?-timepoint_before(1633-07-14, ?-timepoint_before(1609-07-00, ?-timepoint_before(1609-07-19, ?-timepoint_before(1611-03-09, 1636-06-18). 1609-11-00). 1609-07-22). 1611-03-12). In addition, we have deﬁned a predicate for comparing time points in which the year, month and day are the same. We extended the range of conditions under which timepoint same predicate succeeds to include different time measurement units: same month if in one of the time points is measured in month, i.e. the day is 0, we compare that both time points happen in the same month; same year if one of the time points is measured in year, i.e. the month and day are unknown, we compare that both time points happen in the same year. 1 For convenience, the result of this query was trimmed. Chapter 5 Querying the History of Science Ontology 72 timepoint_same(Y-M-D,Y-M-D). timepoint_same(Y-M-0, Y-M-_). timepoint_same(Y-M-_, Y-M-0). timepoint_same(Y-0-0, Y-_-_). timepoint_same(Y-_-_, Y-0-0). The predicate can be exempliﬁed with the corresponding queries which return ’yes’: ?-timepoint_same(1543-05-00, 1543-05-10). ?-timepoint_same(1543-05-10, 1543-05-0). ?-timepoint_same(1543-0-0, 1543-10-10). ?-timepoint_same(1543-10-10, 1543-0-0). In order to infer time relations between events we have deﬁned a number of predicates taken from Allen’s interval algebra. Due to time constraints, we decided to illustrate only a part of the entire set of thirteen relations by considering the ones which were more appropriate for querying and reasoning on our knowledge base. We will deﬁne Prolog predicates for the following relations between events: before, during, overlap, equal time, start together, and end together. According to Allen’s deﬁnition [4], event E1 happens before event E2 when E1 ends before the starting point of E2. This is encoded in the predicatehappen before: happen_before(E1, E2 ) :event_property(end, E1, T_E1e ), event_property(begin, E2, T_E2s ), timepoint_before( T_E1e, T_E2s). For example, we can ﬁnd the events that a particular event (e.g. when Brahe published Nova Stella) preceded: ?-happen_before(d_brahe_publish_novastella, E2 ). E2=d_galileo_write_discourse;E2=d_galileo_publish_assayer; E2=d_kepler_publish_harmonice;E2=d_kepler_publish_astronomia_nova; E2=d_galileo_invent_telescope;E2=d_kepler_brahe_collaboration; E2=d_galileo_investigate_sunspot;E2=d_galileo_observe_sunspot; E2=d_scheiner_sunspot;E2=d_bevis_observe_supernova;E2=d_study_galileo_institution; E2=d_galileo_invent_thermometer;E2=d_galileo_read_elements; E2=d_e_industrial_revolution;E2=d_e_galileo_theory_tides; E2=d_e_kepler_moon_theory;E2=d_e_newton_universal_gravitation; In a similar way, we may deﬁne when events overlap each other. According to Allen’s deﬁnition [4], two events overlap each other if: the start point of E2 is before the end point of E1, the end point of E1 is before the end point of E2, the start point of E1 is before the start point of E2. Consequently, the predicate happen overlap is deﬁned as follows: Chapter 5 Querying the History of Science Ontology 73 happen_overlap(E1,E2):event_property(begin, E1, T_E1s), event_property(end, E1, T_E1e ), event_property(begin, E2, T_E2s), event_property(end, E2, T_E2e), timepoint_before(T_E2s, T_E1e), timepoint_before(T_E1e, T_E2e), timepoint_before(T_E1s, T_E2s). In Allen’s framework, a basic relation named during is deﬁned under the following conditions: the start point of E1 is before the start point of E2, the end point of E1 is before E2. It is important to point out that happen overlap and happen during differ. In overlap, E1 is only partially contained in E2,2 while in during E1 happens within the boundaries of E2. The happen during predicate is deﬁned as follows: happen_during(E1,E2):event_property(begin, E1, T_E1s ), event_property(end, E1, T_E1e), event_property(begin, E2, T_E2s), event_property(end, E2, T_E2e), timepoint_before(T_E2s, T_E1s), timepoint_before(T_E1e, T_E2e). The following query exempliﬁes the happen during predicate: ?-happen_during(d_kepler_publish_harmonice, E2). E2=d_galileo_investigate_sunspot; E2=d_e_scientific_revolution The last predicate selected from Allen’s thirteen interval relations is equals. Applied to events, it is deﬁned as follows: two events happen equal if both have the same start and end points. We have deﬁned the predicate happen equal by using event property and timepoint same: happen_equal(E1,E2):event_property(begin, E1, T_E1s), event_property(end, E1, T_E1e), event_property(begin, E2, T_E2s), event_property(end, E2, T_E2e), \+E1=E2, timepoint_same(T_E1s, T_E2s), timepoint_same(T_E1e, T_E2e),!. The predicate can be used to derive events that happen at the same time as a given event. The following query derives that Galileo invented telescope in the same year when 2 In fact, during is the converse relation of contains. Chapter 5 Querying the History of Science Ontology 74 Kepler published Astronomia Nova. Note that because the second event is measured in year (month and year are 0), the timepoint same predicate will compare the start years and the end years of both events. ?-happen_equal(d_kepler_publish_astronomia_nova, E2). E2=d_galileo_invent_telescope Finally, we added predicates for checking whether two events start together or ﬁnish together. These predicates were not taken from Allen’s framework. However, they were considered appropriate in order to assist the ontology population. happen_same_startpoint(E1,E2):event_property(begin, E1, T_E1s), event_property(begin, E2, T_E2s), T_E1s=T_E2s, \+E1=E2. happen_same_endpoint(E1,E2):event_property(end, E1, T_Es), event_property(end, E2, T_Es1), T_Es=T_Es1, \+E1=E2. The following queries illustrate these predicates: ?-happen_same_startpoint(d_copernicus_publish_derevolutionibus, E2). E2=d_e_scientific_revolution ?-happen_same_endpoint(d_galileo_investigate_sunspot, E2). E2=d_galileo_write_discourse This section deﬁned rules that illustrate the three query modes: Concept-based, Relationbased, and Time-Event related. We have illustrated the application of the rules with example queries. Further application of these rules for composing queries to answer domainrelated questions will be presented in the next section. 5.3 Domain-speciﬁc questions In this section, we exemplify the query modes specifying a number of domain oriented questions which have been subdivided into ﬁve types: Who, What, Where, When, and Combined questions. Each type of question is deﬁned as a display Prolog predicate that prints the result on the screen and is illustrated with example answers derived from the ontology. Chapter 5 Querying the History of Science Ontology 75 5.3.1 Who questions The Who question type is aimed at knowing the person who is the actor of a given action (by action we consider any activity performed to accomplish an objective). The following examples illustrate the application of query mode rules to ﬁnd answers to questions concerning people: Who wrote book B? To create a rule that answers this question, we need to check what was written or published by a person that the relations write and publish can also be expressed via their inverse. who_wrote(B):setof( X, ( inferred_inverse_relation(publish,X,B) ; inferred_inverse_relation(write,X,B) ), All ), showlist(All). The following queries illustrate this rule: ?-who_wrote(’Consequences of the Observations of Capillarity Phenomena’). [’Einstein’] ?-who_wrote(’De nova stella’). [’Brahe’] Who inﬂuenced S? To create a rule that answers this question, we need to take into account that influence is a transitive relation.3 who_influenced(P):setof(P1, inferred_transitive_relation(influence,P1,P),All), showlist( All ). The following query illustrates this rule: ?-who_influenced(’Brahe’). [’Kepler’, ’Copernicus’] Who worked with S? To create a rule that answers this question, we need to take into account that ’work with somebody’ is a symmetrical relations. who_worked_with(P):setof(P1, inferred_symmetrical_relation(’work with somebody’,P,P1),All), showlist( All ). The following query illustrates this rule: ?-who_worked_with(’Kepler’). [’Brahe’, ’Wallenstein’] 3 We made use of the inverse relation of inﬂuence, since this was more appropriate for our needs. Chapter 5 Querying the History of Science Ontology 76 5.3.2 What questions What questions search for objects, and can be illustrated with the following examples: What is the role of person P? The rule utilizes Concept mode to collect all classes of the instance P, to check which of them are roles, and to collect these roles. what_role_has(P):setof( R, ( inferred_instance(P,R), concept_type( R, Type ), inferred_subclass( Type, role ) ), All ), showlist(All). The following queries illustrate this rule: ?-what_role_has(’Galileo’). [scientist] ?-what_role_has(’Aristotle’). [philosopher] ?-what_role_has(’Euclid’). [philosopher, scientist] Which theory extends theory T? The predicate that infers answers to this question takes into account the transitivity of the relation extend. which_theory_extend(T):setof(X, inferred_transitive_relation(extend, T, X), All ), showlist( All ). The following queries exemplify this predicate: ?-which_theory_extend(’galilean theory’). [‘brahe theory’,’ptolemy theory’] ?-which theory_extend(’brahe theory’). [’ptolemy theory’] We can also deﬁne rules for checking what happens at a given time by considering the event relations deﬁned in section 5.2.3. What happened between two time points? To create a rule that answers this question, we will deﬁne a predicate event happened between timepoint to infer the events between two time points by using timepoint before. We can then print this result on the screen with a corresponding display predicate. Chapter 5 Querying the History of Science Ontology 77 event_happened_between_timepoint( T1, T2, E ) :event_property( begin, E, T_Es ), timepoint_before( T1, T_Es ), event_property( end, E, T_Ee ), timepoint_before( T_Ee, T2 ). what_happened_between_timepoint( T1, T2 ) :findall( Happening, ( event_happened_between_timepoint(T1,T2,E), ( ( fact_relation( R, A, B, E ), Happening = [R, A, B, P_e] ) ; ( fact_relation( R, A, E ), Happening = [R, A, P_e] ) ), event_properties_detail( E, P_e ) ), All ), showlist( All ). The following query exempliﬁes the use of this question to ﬁnd out what happened between 1543 and 1750.4 ?-what_happened_between_timepoint(1543-00-00, 1750-00-00). [write,Galileo,Discourse on the Two Sciences,[1633-0-0, 1636-0-0, Italy]] [is written by,The Assayer,Galileo,[1623-0-0, 1623-0-0, Italy]] [publish,Kepler,Harmonice Mundi,[1619-0-0,1619-0-0]] [publish,Kepler,Astronomia Nova,[1609-0-0,1609-0-0,_1647]] [invent,Galileo,telescope,[1609-7-0,1609-11-0]] [work with somebody,Kepler,Brahe,[1600-0-0,1601-0-0,Prague]] [investigate,Galileo,sunspot,[1612-4-0,1636-0-0,Italy]] [observe,Galileo,sunspot,[1611-3-12,1611-3-12,Rome]] [died,Copernicus,[1543-5-25,1543-5-25,Warmia]] [was born,Brahe,[1546-12-14,1546-12-14,Knudstrop]] [observe,Brahe,supernova,[1572-0-0,1572-0-0,Germany]] [observe,Bevis,supernova,[1731-0-0,1731-0-0,England]] [publish,Brahe,De nova stella,[1573-0-0,1573-0-0]] [study,Galileo,[1581-0-0,1585-0-0]] [invent,Galileo,thermometer,[1593-0-0,1593-0-0]] [read,Galileo,Elements,[1583-0-0,1583-0-0]] [explain,galilean theory of tides,tides,[1616-0-0,1616-0-0]] [explain,kepler moon theory,tides,[1609-0-0,1609-0-0] [explain,newton universal gravitation,tides,[1687-0-0,1687-0-0]] What happened in historical period H? To compose a predicate that derives answers to this question, we will use the previous predicate to derive events that happen between the start and end point of a historical period. 4 For convenience, the corresponding result was trimmed. Chapter 5 Querying the History of Science Ontology 78 what_happened_during_historical_period(H) :fact_relation(hold,H,E), event_property(begin,E,T_Es), event_property(end,E,T_Ee), what_happened_between_timepoint( T_Es, T_Ee ). The following query prints the events that happen during the scientiﬁc revolution. ?-what_happened_during_historical_period(’scientific revolution’). d_galileo_write_discourse;d_galileo_publish_assayer;d_kepler_publish_harmonice; d_kepler_publish_astronomia_nova;d_galileo_invent_telescope;d_kepler_brahe_collaboration; d_galileo_investigate_sunspot;d_galileo_observe_sunspot;d_scheiner_sunspot; d_died_copernicus;d_born_brahe;d_observe_brahe_supernova;d_bevis_observe_supernova; d_brahe_publish_novastella;d_study_galileo_institution;d_galileo_invent_thermometer; d_galileo_read_elements;d_e_galileo_theory_tides;d_e_kepler_moon_theory; d_e_newton_universal_gravitation 5.3.3 Where questions Where questions consider the place where a given action occurred. These questions are illustrated with the following examples. Where was P born? The predicate that derives answers to this question writes on the screen the location where somebody was born: where_was_born(P):fact_relation(’was born’, P,E), event_property(location, E, Place), write(Place). This can be exempliﬁed through the following queries: ?-where_was_born(’Brahe’). [Knudstrop] ?-where_was_born(’Ptolemy’). [Alexandria] In a similar way, we can ask for location of different types of events. Where was a phenomenon Ph observed? Answers to this question can be derived with the following predicate: where_observed_phenomenon(Ph):findall(Place, ( fact_relation(observe, _, Ph, E), event_property(location, E, Place) ) , All), showlist(All). Chapter 5 Querying the History of Science Ontology 79 This can be used to infer that the sunspot was observed in Rome (by Galileo) and in Germany (by Scheiner). ?-where_observed_phenomenon(sunspot). [’Rome’, ’Germany’] 5.3.4 When questions When questions are aimed at inferring temporal speciﬁcation with regard to events. The following examples illustrate when questions: When was publication P published? The predicate that derives answers to this question writes on the screen the start point of the event associated with the publishing of P: when_publish(P):setof(Time, ( fact_relation(publish, _, P, E), event_property(begin, E, Time) ) , All), showlist(All). For a given publication, the year when it was published can be derived with the following query: ?- when_publish(’Harmonice Mundi’). 1619-0-0 When did P1 and P2 collaborate? The following predicate checks the time points in between two people work together by considering recursively the interval in which the collaboration takes place: when_collaborate(P1,P2):setof([Interval], (( fact_relation(’work with somebody’, P1,P2,E), Interval=[P1, P2, E, [T_Es, T_Ee]]), event_property(begin, E, T_Es), event_property(end, E, T_Ee)), All), showlist(All). The following query derives the time interval when Kepler and Brahe worked together: ?-when_collaborate(’Kepler’,’Brahe’). [[Kepler,Brahe,d_kepler_brahe_collaboration,[1600-0-0,1601-0-0]]] Chapter 5 Querying the History of Science Ontology 80 5.3.5 Combined Type questions Combined questions involve more than one of the above types of questions. In these cases, we need to derive details about events, for which the predicate event pro perties detail from Time-event mode will be used. We will illustrate combined questions with two examples. I can be noted that, for convenience, the corresponding results were trimmed. When and where was P born? when_where_was_born(P):fact_relation(’was born’, P, E), event_properties_detail(E,P_e), showlist(P_e). The following query exempliﬁes this predicate: ?-when_where_was_born(’Ptolemy’). [was born, Ptolemy, [85-00-00, 85-00-00, Alexandria]] When, where and by whom was phenomenon Ph observed? who_when_where_observed(Ph):findall(P_e, ( fact_relation(observe,P,Ph,E), event_properties_detail(E,P_e) ), All), showlist(All). The following query exempliﬁes this predicate: ?-who_when_where_observed(supernova). [[observe, ’Brahe’, supernova, [1572-00-00, 1572-00-00, Germany]], [[observe, ’Bevis’, supernova, [1731-00-00, 1731-00-00, England]] 5.4 Using queries to verify and expand the ontology According to our methodology, we have considered a set of competency questions for validating the ontology against the original scope and scenario problems. The types of questions previously described have a twofold function: on the one hand they represent an exempliﬁcation of the addressed query modes, on the other hand they have been used as competency questions with the purpose of testing the consistency and correctness of the ontology. When making queries to the ontology, we derived answers which did not appear as expected outputs of our questions. This pointed at errrors in encoding the ontology and missing information. In this section, we will summarize the most recurrent categories of errors and inconsistencies identiﬁed through querying and reasoning upon the ontology. Chapter 5 Querying the History of Science Ontology 81 5.4.1 Syntactical errors These errors refer to inconsistent deﬁnitions. For example, when checking whether Aristo tle is a scientist, the query failed. This was due to inconsistent use of relations in our ontology: facts were encoded both as fact instance of and fact instance. In addition, we identiﬁed the same instances encoded in two different ways. For instance, when trying to infer who invented the telescope the query failed. We realized that the ’galilean telescope’ was recorded either as telescope or as ’three powered galilean telescope’. Examples of misspelled words were also identiﬁed within instantiated relations. For instance, we were inconsistent in naming relations - using inﬁnitive, past tense, present tense, third person. We detected a number of such inconsistencies across the domain. This ended up in difﬁculty to infer facts which we believed were encoded in the ontology. Consequently, we decided to follow a consistent way throughout and to use inﬁnitive tense for naming relations (e.g. we use influence instead of influences or influenced). The ontology was tuned accordingly. 5.4.2 Hierarchical inconsistencies and redundancies Hierarchical inconsistencies were identiﬁed by considering both inferred instances display predicates and types of questions. For example, we detected an inconsistency related to the upper class role. Initially, we considered scientist and philosopher to be direct subclasses of occupational role. However, show inferred instances incorrectly derived that ’Bacon’ was an instance of role. Consequently, we decided to encode that scientist and philosopher were direct subclasses of person but not of role. Occupational role, in fact, is not a superclass of particular role concepts, rather it is a type of concept. This was done by using concept type: concept_type(scientist,’occupational role’). concept_type(philosopher,’occupational role’). This enabled us to consider proprieties of occupational role instead of individuals, avoiding any odd consequence in term of hierarchal structure. 5.4.3 Population of the ontology for answer the queries We also made use of types of questions to identify unpopulated or insufﬁciently populated parts of the ontology for querying and reasoning purposes. These were mostly where more instances needed to be added. When we deﬁned a subset of predicates from Allen’s Chapter 5 Querying the History of Science Ontology 82 algebra, we tried to exemplify and test their use. However, we could not ﬁnd sufﬁcient information for some rules. For example, we could not ﬁnd any existing information to check the happen overlap predicate . Nevertheless, we consider this predicate to be interesting for our domain and we are expected to make further enhancements (discussed in Chapter 6) to populate the ontology for this purpose. With regard to ‘happen’ predicates for comparing events, we initially did not have both begin and end points for each Davidson’s event. This was motivated by the fact that certain events (e.g. publishing a book) do not hold an interval of time, occurring at a precise time point. However, the absence of both temporal boundaries affected our reasoning and did not allow consistently comparing events in which the begin and end points were required. For this reason, when the time of an event was known, we included both a begin and an end point. If the event happened at a single time point, we speciﬁed this time point to be the start, as well as the end of the event. This enabled consistent and powerful reasoning upon the domain. Furthermore, we realized that there were not many symmetrical relations. Some of these were added to ensure that the predicates were correctly tested. 5.5 Summary In this chapter, we have identiﬁed and exempliﬁed a corpus of rules for reasoning upon the domain of History of Science. We deﬁned abstract rules applicable to any domain classiﬁed into three query modes. They concerned the main structures to reason upon: concepts, relations, and time-events. Furthermore, a number of speciﬁc questions illustrating how those modes can be applied into a historical domain were formulated. According to the requirements described in Chapter 2, we mainly aimed at inferring and comparing temporal speciﬁcations, as well as exploring Prolog’s search procedures. It must be noted that while the general rules deﬁned in the query modes are fairly comprehensive, the rules illustrating domain-speciﬁc questions are not exhaustive. We included several representative examples, but there are many more possibilities to compose domain-speciﬁc questions by using rules from the ontology-based query modes. For example, where questions can be expanded to consider not only the places directly encoded but also those which are derived considering the contain relation. In this way, we will be able to infer not only that an event happened at Rome but also that this was in Italy and in Europe. Further expanding on this, we could check what happened in Europe during a particular time. The possibilities are endless. Due to time constraints, we are restricted to several examples. Possible further extension and application is discussed Chapter 5 Querying the History of Science Ontology 83 in the next chapter, where we will give an account of the contributions of our work and discuss its potential enhancements. Chapter 6 Conclusions and Future work 6.1 Synopsis This work has focused on developing a case study in the History of Science domain which illustrates how to conceptualise and reason about a historical domain, by considering a combinative approach to model temporal concepts and relations. We have proposed the development of an ontology-based representation and a reasoning framework by means of three main stages: 1. Elaborating a domain-oriented methodology (presented in Chapter 3) We have developed a three-phase methodology (pre-conceptualization, conceptualization, logical representation and coding), for enabling us to conceptualise and encode the domain, by identifying the main categories and relations. With regard to domain-oriented application of existing methodologies (Chapter 2), our methodology aims at supporting the role of the domain expert who in this work has acted as knowledge engineer as well. It, also, gave us a systematic way to identify temporal concepts and relations in which time dimensions have to be added (Chapter 4). 2. Identifying and modeling time-embedded relations (presented in Chapter 4) We have proposed an approach for identifying and representing time in relations by considering the distinction between type and token and an adaptation of Davidson’s 84 Chapter 6 Conclusions and Future work 85 theory of events. Such an approach considers existing theoretical and ontologybased representations (Chapter 2) and it can potentially be applied to any historical domains for which we have outlined general characteristics and challenges (Chapter 2). 3. Querying and reasoning on the domain (presented in Chapter 5) We have deﬁned an ontology-based rules framework for our domain. A corpus of abstract rules were deﬁned and exempliﬁed by considering speciﬁc questions in the History of Science domain. Such a framework integrates a sub-set of predicates from Allen’s algebra for reasoning about temporal relations. 6.2 Main contributions of this thesis The primary contribution of this work was to combine Davidson’s theory of events and Allen’s interval algebra into a uniﬁed framework for adding and comparing time-occurrences. Such a combination ensured a consistent result and it looks promising as a basis for further enhancements. Davidson’s theory enabled us to easily associate proprieties to events without encoding a unmanageable number of instantiated relations, while Allen’s relations assisted us deﬁning predicates for comparing time in events. The distinction between type and token for time inclusion enabled us to systematically apply the token reiﬁcation approach of events. Such a distinction was neat and it enables straightforward decisions with regard to time inclusion. With regard to the methodological aspects, our work is aimed at scaling and adapting features of existing methodologies, mostly inspired by enterprise modeling and software engineering [88] [39] [87] [86], for building a history ontology which is conceptualized and formalized by a domain expert. Our case study challenged the role of the expert, bridging the gap between non technical and technical expertises . 6.3 Limitations In this section, we outline the limitations of our work, by considering methodological and time modeling aspects. As follows, we sketched out the most crucial limitations we have identiﬁed during our work: Chapter 6 • Methodology scope Conclusions and Future work 86 Our methodology addressed a speciﬁc case in which the domain expert is involved in the logical and encoding phase as well. Nevertheless, the development of an ontology is commonly performed by teams in which a clear distinction between the roles of expert and knowledge engineer is required. The scope of methodology applicability has been limited by the fact that it was deﬁned ad hoc for our case study. Thus, the criterion of applicability would likely fail using our methodology for diverse case studies. • Ontology population Some parts of the ontology were not enough populated with facts and relations. In some cases, instances are missing, in others we deﬁned relations in our data model, but we did not exemplify them by considering their instantiated counterpart. Some predicates deﬁned in Chapter 5 (e.g. happen overlap) were tested by considering only a number of test examples. • Reasoning limitation: predicates and domain-speciﬁc questions Our rules might have much more elegant formulation. For instance, inferred relation predicates as they are can be better deﬁned by considering a second-level recursion which derives all transitive, inverse and symmetrical clauses. Furthermore, we point at the need of more generic query modes which can combine speciﬁc rules automatically. In addition, the domain oriented questions (competency questions) were not identiﬁed by means of a systematic approach. However, they were formulated for the sake of illustrating the domain and an exhaustive account of them was our of the scope of this work. • Ontology validation Methodological limitations relate to the ontology validation. An appropriate validation of the ontology is missing. Due to time constraint, we did not consider to validate our knowledge model against external domain experts. However, in order to expose the ontology to external validation, we need to make it available via a user friendly interface. Such a limitation can be tackled by addressing the enhancement which considers a possible application of the ontology scenario (e.g. web-based prototype), as describe in Section 6.3.1. Finally, it should be noted that we exposed our ontology to an internal validation performed by two knowledge engineers which were primarily involved in the logical phase. Chapter 6 Conclusions and Future work 87 6.3.1 Future work In this section we outline some possible enhancements to be addressed as a part of a future work: • Application of the ontology scenario A potential enhancement might consider the implementation of a user interface to query and display the ontology. This can be done, for instance, through the use of a CGI Prolog library which dynamically generates HTML documents by directly interrogating the ontology. Such an application would accomplish the scope of validate the ontology, as mentioned above. • Dissemination and interoperability of our data model For the sake of dissemination and interoperability, a number of existing libraries1 can be used to convert Prolog terms into OWL abstract syntax. This conversion can be employed to verify whether other ontology languages (e.g. OWL) are enough powerful and expressive to represent our model. By considering the logical form of our rules, they can be translated into an alternative query language like SPARQL.2 • Allowing query composition With the purpose of automatically combining different queries , we can modify the predicates for domain speciﬁc questions, deﬁned in Chapter 5, by adding an extra argument to capture the output list. In this way, the domain questions can be used not only to list results of competency queries, but also for further reasoning upon the domain. For instance the query who when where observed(supernova), as deﬁned in page 81, will integrate the second parameter P e, which is a list that all event properties in form of location and temporal speciﬁcations. P e can then be used in more elaborated questions. • Expand time representation We might consider our time representation to be able to handle imprecise temporal boundaries. For instance: OWL Prolog Library documentation:http://www.semanticweb.gr/TheaOWLLib/Thea OWL Lib.pdf (Visited, August 2006). 2 SPARQL query language: http://www.w3.org/TR/rdf-sparql-query/ (Visited, September 2006). 1 Thea Chapter 6 Conclusions and Future work 88 Applying interval-based approach for historical periods We might fuzzily represent boundaries of historical periods by considering an intervalbased encoding. Furthermore, we might add enhancements which address: Improving accuracy in dating mechanism We might implement rules for distinguishing between events which occur before and after Christ. This enables us to efﬁciently reason about different events based on different dating conventions (e.g we can deﬁne an event which occurred in 85 BC and one in 85 AD). Deﬁning geopolitical naming convention We might want to capture changes in geopolitical naming. Dealing with historical domains, we might require to encode cities which have changed name or that have became part of a different geographical area. • Consider other challenges Representing uncertainty and vagueness in historical domains can be addressed as future work. Once, we achieve more soundness and expressiveness with regard to our time representation, we can deploy this to reach different levels of granularity. Bibliography [1] AA.VV. Teaching the History of Science. Blackwell, 1989. [2] P. Achinstein. Concepts of Science. The Johns Hopkins Press, 1968. [3] J.F. Allen. Toward a general theory of action and time. Artiﬁcial Intelligence, 23:pages 123–154, 1984. [4] J.F. Allen. Time and time again: the many ways to represent time. International Journal of Intelligent Systems, 6:pages 341–355, 1991. [5] J.F. Allen and G. Ferguson. Actions and events in interval temporal logic. Journal of Logic and Computation, 4(5):pages 531–579, 1994. [6] J.F. Allen and P.J. Hayes. Moments and points in an interval-based temporal logic. Computational Intelligence, 5(4):pages 225–238, 1989. [7] B. Bennett and A.P. Galton. A unifying semantics for time and events. Artiﬁcial Intelligence, 153(1-2):pages 13–48, 2004. [8] R.I. Binnick. Time and the Verb: A guide to tense and Aspect. Oxford University Press, 1991. [9] E. Burns and S. Law. Philosophy for AS and A2. Routledge, 2004. [10] W.F. Bynum, Browne E.J., and R. Porter. Dictionary of the History of Science. The Macmillan Press LTD, 1981. [11] A.F. Chalemers. What is this thing called Science. Open University Press, 1999. [12] I.B. Cohen. Revolution in Science. Harvard Univ, 1985. [13] B. Comrie. Aspect: An Introduction to the Study of Verbal Aspect and Related problems. Cambridge University Press, 1976. 89 BIBLIOGRAPHY 90 [14] O. Corcho, M. Fernandez-Lopez, and A.G. Perez. Methodologies, tools and languages for building ontologies. where is their meeting point? Data and Knowledge Engineering, (46):pages 41–64, 2003. [15] G. Couvalis. The Philosophy of Science. SAGE Publication, 1997. [16] D. Davidson. The logical form of action sentences. In N. Rescher, editor, The Logic of Decision and Action. University of Pitsburgh Press, 1967. [17] D. Davidson. Truth and Meaning, pages 155–170. Oxford University Press, 1990. [18] R. De Hoog. Methodologies for Building Knowledge Based Systems Achievements and Prospects. CRC Press, 1998. [19] R. Deswarte and J. Oosthoek. Creating a history ontology: The challenges and criteria of building a humanities ontology. In Conference National Centre for eSocial Science 2005 (NCeSS), 2005. [20] E.J. Dijksterhuis. The Mechanization of the World Picture. Princepton University Press, 1986. [21] P.T. Durbin. Dictionary of Concepts in the Philosophy of Science. Greenwood Press:New York, 1988. [22] C. Matuszek et al. An introduction to the syntax and content of cyc. In Proceedings of the 2006 AAAI Spring Symposium on Formalizing and Compiling Background Knowledge and Its Applications to Knowledge Representation and Question Answering, 2006. [23] J.C. Arpirez et al. Reference ontology and (onto)2 agent: The ontology yellow pages. In Workshop on Application of Ontologies and Problem-Solving Methods, European Conference on Artiﬁcial Intelligence (ECAI’98), 1998. [24] J.H. Gennari et al. The evolution of protege: an environment for knowledge-based systems development. Human-Computer Studies, pages 89–123, 2003. [25] K. Kovacs et al. A methodology for building conceptual domain ontologies. Technical report, Ordnance Survey Research and Innovation, 2006. [26] M. Fernandez-Lopez et al. Building a chemical ontology using methontology and the ontology design environment. IEEE Intelligent Systems, 14(1):pages 37–46, 1999. BIBLIOGRAPHY 91 [27] Y. Sure et al. Ontoedit: collaborative ontology engineering for the semantic web. Lectures Notes in Computer Science, pages 221–235, 2002. [28] R. Kusters F. Baader and Frank Wolter. Extensions to description logics, pages 219–261. Cambridge University Press, New York, NY, USA, 2003. [29] M. Fernandez-Lopez. Overview of the methodologies for building ontologies. In IJCAI99 Workshop on Ontologies and Problem-Solving Methods: Lessons Learned and Future Trends, 1999. [30] M. Fernandez-Lopez and A. Gomez-Perez. Overview and analysis of methodologies for building ontologies. Knowledge Engineering Review, 17(2):pages 129–156, 2002. [31] M.S. Fox. The tove project: Towards a common-sense model of the enterprise. Technical report, Enterprise Integration Laboratory Technical Report, 1992. [32] W. Frawley. Linguistics Semantics. Lawrence Erlbaum, 1992. [33] A.P. Galton. Temporal Logics and their Applications. Academic Press:London, 1987. [34] A.P. Galton. Reiﬁed temporal theories and how to unreify them. In Proceeding of IJCAI’91, pages 1177–1183, 1991. [35] S. Glasbey. Event structure, punctuality, and when. Natural Language Semantics, 12(2):pages 191–211, 2004. [36] A.I. Goldman. The individuation of action. Journal of Philosophy, 68,:pages 761– 774, 1971. [37] A. Gomez-Perez, M. Fernandez-Lopez, and O. Corcho. Methodologies and Methods for building ontologies, pages 107–197. Springer Verlag, 2003. [38] A. Gomez-Perez, M. Fernandez-Lopez, and O. Corcho. Theoretical Foundations of Ontologies, pages 1–45. Springer Verlag, 2003. [39] A. Gomez-Perez, M. Fernandez-Lopez, and N. Juristo. Methontology: from ontological art toward ontological engineering. In Proceedings of the AAAI97 Spring Symposium Series on Ontological Engineering, pages 33–40, 1997. [40] A. Gomez-Perez and M.D. Rojas. Ontological reengineering and reuse. In European Knowledge Acquisition Workshop (EKAW), 1999. BIBLIOGRAPHY 92 [41] T.R. Gruber. A translation approach to portable ontology speciﬁcations. Knowledge Acquisition, 5(2):pages 199–220., 1993. [42] T.R. Gruber. Toward principles for the design of ontologies used for knowledge sharing. International Journal of Human Computer Studies, 43(5-6):pages 907– 928., 1995. [43] M. Gruninger and M. Fox. Methodology for the design and evaluation of ontologies. In IJCAI’95, Workshop on Basic Ontological Issues in Knowledge Sharing, 1995. [44] N. Guarino. Formal ontology and information systems. In Proceedings of the 1st International Conference on Formal Ontologies in Information Systems, FOIS’98, pages 3–15. IOS Press, 1998. [45] N. Guarino and C. Welty. Evaluating ontological decisions with ONTOCLEAN. Communications of the ACM, 45(2):pages 61–65, 2002. [46] R. Harr. The Philosophies of Science. Oxford University Press, 1985. [47] J.K. Heilbron. The Oxford Companion to the History of Modern Science. Oxford University Press:New York, 2003. [48] J.K. Heilbron. The Oxford Guide to the History of Physics and Astronomy. Oxford University Press:New York, 2005. [49] J.R. Hobbs and F. Pan. An ontology of time for the semantic web. ACM Transactions on Asian Language Processing (TALIP): Special issue on Temporal Information Processing, 3(1):pages 66–85, 2004. [50] IEEE. IEEE Standard Glossary of Software Engineering Terminology (IEEE Std 610.121990). IEEE Computer Society, 1990. [51] IEEE. IEEE Standard for Developing Software Life Cycle Processes. IEEE Computer Society, 1996. [52] M. Jarrar. Towards the notion of gloss, and the adoption of linguistic resources in formal ontology engineering. In Proceedings of the 15th international conference on World Wide Web (WWW2006), pages 497–503. ACM Press, New York, NY, 2006. [53] J. Kim. Events as Property Exempliﬁcations, pages 159–177. Reidel, Dordrecht, 1971. BIBLIOGRAPHY 93 [54] R.A Kowalski and M.J. Sergot. A logic-based calculus of events. New Generation Computing, 4:pages 67–95, 1986. [55] K. Kripke. Semantical considerations on modal logic. Acta Philosophica Fennica, 16:pages 83–94, 1963. [56] G. Leibniz. The Leibniz-Clarke Correspondence: Together with Extracts from Newton Principia and Optics. Manchester University Press, 1956. [57] D.B. Lenat and R.V. Guha. Building large knowledge-based systems: representation and inference in the Cyc project. Addison-Wesley, 1990. [58] E. Lepore and K. Ludwig. The Essential Davidson. Oxford University Press:Oxford, 2006. [59] E. Lepore and B. Smith. The Oxford Handbook of Philosophy of Language. Oxford University Press: Oxford, 2000. [60] Y. Lin. Two theories of time. Journal of Applied Non-Classical Logics, 1(1):pages 37–63, 1991. [61] J. Losee. Theories of Scientiﬁc Progress. Routledge, 2004. [62] J. Lyons. Semantics. Cambridge:Cambridge University Press, 1977. [63] J. Lyons. Linguistic Semantics: An Introduction. Cambridge:Cambridge University Press, 1995. [64] J. Malpas. Donald davidson. In E.N. Zalta, editor, The Stanford Encyclopedia of Philosophy. 2005. [65] J. McCarthy and P.J. Hayes. Some philosophical problems from the standpoint of artiﬁcial intelligence. In B. Meltzer and D. Michie, editors, Machine Intelligence, pages 463–502. Edinburgh University Press, 1969. [66] D.V. McDermott. A temporal logic for reasoning about processes and plans. Cognitive Science, 6:pages 101–155, 1982. [67] H. Mizen, C. Dolbear, and G. Hart. Ontology ontogeny: Understanding how an ontology is created and developed. In First International Conference on Geospatial Semantic (Geos’05), pages 15–29, 2003. BIBLIOGRAPHY 94 [68] B. Motik, U. Hustadt, and U. Sattler. Reasoning in description logics with a concrete domain in the framework of resolution. In Proceedings of the 16th European Conference on Artiﬁcial Intelligence (ECAI 2004), pages 353–357, 2004. [69] B. Motik and U. Sattler. Practical dl reasoning over large aboxes with kaon 2. 2006. [70] A.P.D. Mourelatos. Events, processes and states. 2(1):pages 415–434, 1978. Linguistic and Philosopy, [71] J. et al. Mylopoulos. Telos: Representing knowledge about information systems. Information Systems, 8(4):pages 325–362, 1990. [72] G. Nagypal. History ontology building: The technical view. In Proceedings of the XVI international conference of the Association for History and Computing (AHC 2005), pages 207–214., Royal Netherlands Academy of Arts and Sciences, Amsterdam, 2005. [73] G. Nagypal, R. Deswarte, and J. Oosthoek. Applying the semantic web: The vicodi experience in creating visual contexualization for history. Literary and Linguistic Computing, 20(3):pages 327–349., 2005. [74] I. Newton. Philosophiae Naturalis Principia Mathematica: The third edition. Cambridge University Press, 1972. [75] F. Pan and J.R. Hobbs. Time owl-s. In Proceedings of the AAAI Spring Symposium on Semantic Web Services, pages 29–36, 2004. [76] F Pan and J.R. Hobbs. Temporal aggregates in owl for the semantic web. In Proceedings of the AAAI Fall Symposium on Agents and the Semantic Web. AAAI Press, 2005. [77] C. S. Peirce. Collected Papers. Harvard University Press, 1933. [78] R. Porter and M. Teich. The Scientiﬁc Revolution in National Context. Cambridge University Press, 1992. [79] A. Prior. Time and Modality. Clarendon Press:Oxford, 1957. [80] W. V. Quine. On the theory of types. in Journal of Symbolic Logic, pages pages 125–139, 1938. [81] W. V. Quine. On What There is. Harper and Row, New York, 1953. BIBLIOGRAPHY 95 [82] G. Sandstrom. When-clauses and the temporal interpretation of narrative discourse. Technical report, Department of General Linguistics, University of Ume˙ , 1993. a [83] J.F. Sowa. Knowledge Representation: Logical, Philosophical, and Computational Foundations. Brooks Cole Publishing Co:Paciﬁc Grove, CA, 1999. [84] R. Taton, W. Curtis, and M. Hoskin. The General History of Astronomy. Cambridge University Press, 1995. [85] M. Thompson. Philosphy of Science. Mc Graw Hill, 2005. [86] U. Uschold and M. Gruninger. Ontologies: Principles, methods and applications. Knowledge Engineering Review, 11(2), 1996. [87] U. Uschold and M. King. Towards a methodology for building ontologies. In Workshop on Basic Ontological Issues in Knowledge Sharing, held in conduction with IJCAI-95, 1995. [88] L. Iverson V. Mirzaee and B. Hamidzadeh. Towards ontological modelling of historical documents. In Protege Conference 2004, 2004. [89] G. van Heijst, A. Schreiber, and B.J.Wielinga. Using explicit ontologies in kbs development. International Journal of Human-Computer Studies, 45(183):pages 183–292, 1997. [90] Z. Vendler. Verbs and times. The Philosophical Review, pages 143–160, 1957. [91] Z. Vendler. Linguistic in Philosophy. Cornell University Press:New York, 1967. [92] L. Vila. A survey on temporal reasoning in artiﬁcial intelligence. AI Communications, 7(1):pages 4–28, 1994. [93] L. Vila and H. Reichgelt. The token reiﬁcation approach to temporal reasoning. Artiﬁcial Intelligence, 83(1):pages 59–74, 1996. [94] P. Vossen. Eurowordnet: a multilingual database for information retrieval. In In: Proceedings of the DELOS workshop on Cross-language Information Retrieval, 1997. [95] L. Wetzel. Types and Tokens: An Essay on Universals. Boston, Massacusetts: MIT Press, 2006. BIBLIOGRAPHY 96 [96] K. Wilcox. Semantic web technologies in cultural history: The state of the art. Technical report, Azusa Paciﬁc University, University Libraries, 2005. [97] J. Zhang. Semantic web: Mapping metadata is more: Making more use out of existing metadata from digital libraries using semantic web technologies. Mres thesis, University of Groningen, 2006. [98] Q. Zhou and R. Fikes. A reusable time ontology. Technical report, Knowledge Stanford Laboratory, 2001. Appendix A Expanded queries A.1 Example queries The extended list of queries includes additional examples of Concept and Relation-based display predicates, as described in Chapter 5. These predicates will enable the reader to explore the structure of the History of Science ontology. The entire ontology is available through the following URL: http://www.comp.leeds.ac.uk/ilaria/hiso.pl A.1.1 Concept Mode queries: additional examples show inferred subclasses predicate ?- show_inferred_subclasses. inferred_subclass(abductive argument,mode of reasoning) inferred_subclass(academic organizaton,group of people) inferred_subclass(analogical argument,mode of reasoning) inferred_subclass(analytic,philophical doctrine) inferred_subclass(anatomist,person) inferred_subclass(anatomist,scientist) inferred_subclass(ancient,historical period) inferred_subclass(ancient,time) inferred_subclass(anomaly,event) inferred_subclass(anomaly,observation) inferred_subclass(article,document) inferred_subclass(astro area,place) inferred_subclass(astrologer,person) inferred_subclass(astrologer,philosopher) inferred_subclass(astronomer,person) 97 Appendix A Expanded queries 98 inferred_subclass(astronomer,scientist) inferred_subclass(book,document) inferred_subclass(character role,role) inferred_subclass(chemist,person) inferred_subclass(chemist,scientist) inferred_subclass(city,geopolitical area) inferred_subclass(city,place) inferred_subclass(classical method,method) inferred_subclass(cosmologist,person) inferred_subclass(cosmologist,philosopher) inferred_subclass(deductive argument,mode of reasoning) inferred_subclass(discoverer,person) inferred_subclass(discoverer,scientist) inferred_subclass(discovery,event) inferred_subclass(dualism,doctrine) inferred_subclass(dualism,philosophical doctrine) inferred_subclass(early modern,historical period) inferred_subclass(early modern,time) inferred_subclass(empire,geopolitical area) inferred_subclass(empire,place) inferred_subclass(epistemologist,person) inferred_subclass(epistemologist,philosopher) inferred_subclass(experiment,event) inferred_subclass(galaxy,astro area) inferred_subclass(galaxy,place) inferred_subclass(geographical area,place) inferred_subclass(geopolitical area,place) inferred_subclass(historical period,time) inferred_subclass(humanities,field of study) inferred_subclass(hypothesis,model) inferred_subclass(idealism,doctrine) inferred_subclass(idealism,philosophical doctrine) inferred_subclass(in book,document) inferred_subclass(in proceeding,document) inferred_subclass(inductive argument,mode of reasoning) inferred_subclass(intuitionism,doctrine) inferred_subclass(intuitionism,philosophical doctrine) inferred_subclass(invention,event) inferred_subclass(inventor,person) inferred_subclass(inventor,scientist) inferred_subclass(journal,document) inferred_subclass(kingdom,geopolitical area) inferred_subclass(kingdom,place) inferred_subclass(law,model) inferred_subclass(logician,person) inferred_subclass(logician,philosopher) inferred_subclass(manual,document) inferred_subclass(manuscript,document) inferred_subclass(materialism,doctrine) inferred_subclass(materialism,philosophical doctrine) inferred_subclass(mathematician,person) inferred_subclass(mathematician,scientist) inferred_subclass(mecanicism,doctrine) inferred_subclass(mecanicism,philosophical doctrine) inferred_subclass(mentalism,doctrine) Appendix A Expanded queries 99 inferred_subclass(mentalism,philosophical doctrine) inferred_subclass(metaphysical cosmologist,cosmologist) inferred_subclass(metaphysical cosmologist,person) inferred_subclass(metaphysical cosmologist,philosopher) inferred_subclass(methaphysician,person) inferred_subclass(methaphysician,philosopher) inferred_subclass(middle ages,historical period) inferred_subclass(middle ages,time) inferred_subclass(modern,historical period) inferred_subclass(modern,time) inferred_subclass(monism,doctrine) inferred_subclass(monism,philosophical doctrine) inferred_subclass(natural philosopher,person) inferred_subclass(natural philosopher,philosopher) inferred_subclass(naturalist,person) inferred_subclass(naturalist,scientist) inferred_subclass(non academic organization,group of people) inferred_subclass(observation,event) inferred_subclass(occupational role,role) inferred_subclass(phenomenology,doctrine) inferred_subclass(phenomenology,philosophical doctrine) inferred_subclass(philosopher,person) inferred_subclass(philosophical belief,belief) inferred_subclass(philosophical doctrine,doctrine) inferred_subclass(philosophical school,group of people) inferred_subclass(physician,person) inferred_subclass(physician,scientist) inferred_subclass(physicist,person) inferred_subclass(physicist,scientist) inferred_subclass(planet,astro area) inferred_subclass(planet,place) inferred_subclass(pragmatism,doctrine) inferred_subclass(pragmatism,philosophical doctrine) inferred_subclass(proceeding,document) inferred_subclass(province,geopolitical area) inferred_subclass(province,place) inferred_subclass(pseudo science,field of study) inferred_subclass(psycological belief,belief) inferred_subclass(rationalism,doctrine) inferred_subclass(rationalism,philosophical doctrine) inferred_subclass(realism,doctrine) inferred_subclass(realism,philosophical doctrine) inferred_subclass(religious belief,belief) inferred_subclass(science,field of study) inferred_subclass(scientific doctrine,doctrine) inferred_subclass(scientific method,method) inferred_subclass(scientific school,group of people) inferred_subclass(scientist,person) inferred_subclass(star,astro area) inferred_subclass(star,place) inferred_subclass(state,geopolitical area) inferred_subclass(state,place) inferred_subclass(technical report,document) inferred_subclass(theory,model) inferred_subclass(thesis,document) Appendix A Expanded queries 100 inferred_subclass(time interval,time) inferred_subclass(time point,time) inferred_subclass(time unit,time) inferred_subclass(university,academic organization) inferred_subclass(web page,document) show inferred subclass of (historical period). ?- show_inferred_subclasses_of(’historical period’). ancient early modern middle ages modern show inferred subclasses of(doctrine) ?- show_inferred_subclasses_of(doctrine). dualism idealism intuitionism materialism mecanicism mentalism monism phenomenology philosophical doctrine pragmatism rationalism realism scientific doctrine show indirect subclasses predicate ?- show_indirect_subclasses. indirect_subclass(anatomist,person) indirect_subclass(ancient,time) indirect_subclass(anomaly,event) indirect_subclass(astrologer,person) indirect_subclass(astronomer,person) indirect_subclass(chemist,person) indirect_subclass(city,place) indirect_subclass(cosmologist,person) indirect_subclass(discoverer,person) indirect_subclass(dualism,doctrine) indirect_subclass(early modern,time) indirect_subclass(empire,place) indirect_subclass(epistemologist,person) indirect_subclass(galaxy,place) indirect_subclass(idealism,doctrine) indirect_subclass(intuitionism,doctrine) indirect_subclass(inventor,person) indirect_subclass(kingdom,place) indirect_subclass(logician,person) Appendix A Expanded queries 101 inferred_subclass(mentalism,philosophical doctrine) inferred_subclass(metaphysical cosmologist,cosmologist) inferred_subclass(metaphysical cosmologist,person) inferred_subclass(metaphysical cosmologist,philosopher) inferred_subclass(methaphysician,person) inferred_subclass(methaphysician,philosopher) inferred_subclass(middle ages,historical period) inferred_subclass(middle ages,time) inferred_subclass(modern,historical period) inferred_subclass(modern,time) inferred_subclass(monism,doctrine) inferred_subclass(monism,philosophical doctrine) inferred_subclass(natural philosopher,person) inferred_subclass(natural philosopher,philosopher) inferred_subclass(naturalist,person) inferred_subclass(naturalist,scientist) inferred_subclass(non academic organization,group of people) inferred_subclass(observation,event) inferred_subclass(occupational role,role) inferred_subclass(phenomenology,doctrine) inferred_subclass(phenomenology,philosophical doctrine) inferred_subclass(philosopher,person) inferred_subclass(philosophical belief,belief) inferred_subclass(philosophical doctrine,doctrine) inferred_subclass(philosophical school,group of people) inferred_subclass(physician,person) inferred_subclass(physician,scientist) inferred_subclass(physicist,person) inferred_subclass(physicist,scientist) inferred_subclass(planet,astro area) inferred_subclass(planet,place) inferred_subclass(pragmatism,doctrine) inferred_subclass(pragmatism,philosophical doctrine) inferred_subclass(proceeding,document) inferred_subclass(province,geopolitical area) inferred_subclass(province,place) inferred_subclass(pseudo science,field of study) inferred_subclass(psycological belief,belief) inferred_subclass(rationalism,doctrine) inferred_subclass(rationalism,philosophical doctrine) inferred_subclass(realism,doctrine) inferred_subclass(realism,philosophical doctrine) inferred_subclass(religious belief,belief) inferred_subclass(science,field of study) inferred_subclass(scientific doctrine,doctrine) inferred_subclass(scientific method,method) inferred_subclass(scientific school,group of people) inferred_subclass(scientist,person) inferred_subclass(star,astro area) inferred_subclass(star,place) inferred_subclass(state,geopolitical area) inferred_subclass(state,place) inferred_subclass(technical report,document) inferred_subclass(theory,model) inferred_subclass(thesis,document) Appendix A Expanded queries 102 subclass_path(idealism,[philosophical doctrine],doctrine) subclass_path(intuitionism,[philosophical doctrine],doctrine) subclass_path(inventor,[scientist],person) subclass_path(kingdom,[geopolitical area],place) subclass_path(logician,[philosopher],person) subclass_path(materialism,[philosophical doctrine],doctrine) subclass_path(mathematician,[scientist],person) subclass_path(mecanicism,[philosophical doctrine],doctrine) subclass_path(mentalism,[philosophical doctrine],doctrine) subclass_path(metaphysical cosmologist,[cosmologist],philosopher) subclass_path(metaphysical cosmologist,[cosmologist,philosopher],person) subclass_path(methaphysician,[philosopher],person) subclass_path(middle ages,[historical period],time) subclass_path(modern,[historical period],time) subclass_path(monism,[philosophical doctrine],doctrine) subclass_path(natural philosopher,[philosopher],person) subclass_path(naturalist,[scientist],person) subclass_path(phenomenology,[philosophical doctrine],doctrine) subclass_path(physician,[scientist],person) subclass_path(physicist,[scientist],person) subclass_path(planet,[astro area],place) subclass_path(pragmatism,[philosophical doctrine],doctrine) subclass_path(province,[geopolitical area],place) subclass_path(rationalism,[philosophical doctrine],doctrine) subclass_path(realism,[philosophical doctrine],doctrine) subclass_path(star,[astro area],place) subclass_path(state,[geopolitical area],place) show inferred instance predicate show_inferred_instances. inferred_instance(Galileo,person) inferred_instance(Galileo,scientist) inferred_instance(Germany,geopolitical area) inferred_instance(Germany,place) inferred_instance(Germany,state) inferred_instance(Gilbert,natural philosopher) inferred_instance(Gilbert,person) inferred_instance(Gilbert,philosopher) inferred_instance(Graz,city) inferred_instance(Graz,geopolitical area) inferred_instance(Graz,place) inferred_instance(Halley,astronomer) inferred_instance(Halley,person) inferred_instance(Halley,scientist) inferred_instance(Harmonice Mundi,book) inferred_instance(Harmonice Mundi,document) inferred_instance(Harvey,person) inferred_instance(Harvey,physician) inferred_instance(Harvey,scientist) inferred_instance(Hipparcus,astronomer) inferred_instance(Hipparcus,mathematician) inferred_instance(Hipparcus,person) inferred_instance(Hipparcus,scientist) Appendix A Expanded queries 103 inferred_instance(Hobbes,person) inferred_instance(Hobbes,philosopher) inferred_instance(Huygens,astronomer) inferred_instance(Huygens,mathematician) inferred_instance(Huygens,person) inferred_instance(Huygens,physicist) inferred_instance(Huygens,scientist) inferred_instance(Italy,geopolitical area) inferred_instance(Italy,place) inferred_instance(Italy,state) inferred_instance(Kepler,astronomer) inferred_instance(Kepler,mathematician) inferred_instance(Kepler,person) inferred_instance(Kepler,scientist) inferred_instance(Kingdom of Poland,geopolitical area) inferred_instance(Kingdom of Poland,kingdom) inferred_instance(Kingdom of Poland,place) inferred_instance(Knudstrop,city) inferred_instance(Knudstrop,geopolitical area) inferred_instance(Knudstrop,place) inferred_instance(Leibniz,person) inferred_instance(Leibniz,philosopher) inferred_instance(Methaphysics,book) inferred_instance(Methaphysics,document) inferred_instance(Newton,astronomer) inferred_instance(Newton,mathematician) inferred_instance(Newton,person) inferred_instance(Newton,physicist) inferred_instance(Newton,scientist) inferred_instance(Nicaea,city) inferred_instance(Nicaea,geopolitical area) inferred_instance(Nicaea,place) inferred_instance(On the Magnet and Magnetic Bodies,book) inferred_instance(On the Magnet and Magnetic Bodies,document) inferred_instance(Pascal,person) inferred_instance(Pascal,philosopher) inferred_instance(Philosophiae Naturalis Principia Mathematica,book) inferred_instance(Philosophiae Naturalis Principia Mathematica,document) inferred_instance(Pisa,city) inferred_instance(Pisa,geopolitical area) inferred_instance(Pisa,place) inferred_instance(Prague,city) inferred_instance(Prague,geopolitical area) inferred_instance(Prague,place) inferred_instance(Ptolemy,astrologer) inferred_instance(Ptolemy,astronomer) inferred_instance(Ptolemy,mathematician) inferred_instance(Ptolemy,person) inferred_instance(Ptolemy,philosopher) inferred_instance(Ptolemy,scientist) inferred_instance(Roger Bacon,person) inferred_instance(Roger Bacon,philosopher) inferred_instance(Roman Empire period,ancient) inferred_instance(Roman Empire period,historical period) inferred_instance(Roman Empire period,time) Appendix A Expanded queries 104 inferred_instance(Royal Prussia,geopolitical area) inferred_instance(Royal Prussia,place) inferred_instance(Royal Prussia,province) inferred_instance(Scania,geopolitical area) inferred_instance(Scania,place) inferred_instance(Scania,province) inferred_instance(Scheiner,astronomer) inferred_instance(Scheiner,discoverer) inferred_instance(Scheiner,person) inferred_instance(Scheiner,physicist) inferred_instance(Scheiner,scientist) inferred_instance(Sidereus Nuncius,book) inferred_instance(Sidereus Nuncius,document) inferred_instance(University of Graz,academic organization) inferred_instance(University of Graz,university) inferred_instance(Vesalius,anatomist) inferred_instance(Vesalius,person) inferred_instance(Vesalius,scientist) inferred_instance(Warmia,city) inferred_instance(Warmia,geopolitical area) inferred_instance(Warmia,place) inferred_instance(aristotelian method,classical method) inferred_instance(aristotelian method,method) inferred_instance(arts,field of study) inferred_instance(arts,humanities) inferred_instance(astrology,field of study) inferred_instance(astrology,pseudo science) inferred_instance(astronomy,field of study) inferred_instance(astronomy,science) inferred_instance(biology,field of study) inferred_instance(biology,science) inferred_instance(chemistry,field of study) inferred_instance(chemistry,science) inferred_instance(copernicus theory,model) inferred_instance(copernicus theory,theory) inferred_instance(crystalline spheres,belief) inferred_instance(crystalline spheres,hypothesis) inferred_instance(crystalline spheres,model) inferred_instance(crystalline spheres,philosophical belief) inferred_instance(deductive method,classical method) inferred_instance(deductive method,method) inferred_instance(early middle ages,historical period) inferred_instance(early middle ages,middle ages) inferred_instance(early middle ages,time) inferred_instance(electromagnetism,phenomenon) inferred_instance(electron discovery,discovery) inferred_instance(electron discovery,event) inferred_instance(evolutionary theory,model) inferred_instance(evolutionary theory,theory) inferred_instance(galilean theory,model) inferred_instance(galilean theory,theory) inferred_instance(galilean theory of tides,model) inferred_instance(galilean theory of tides,theory) inferred_instance(generalization,inductive argument) inferred_instance(generalization,mode of reasoning) Appendix A Expanded queries 105 inferred_instance(geocentric theory,model) inferred_instance(geocentric theory,theory) inferred_instance(gravity,phenomenon) inferred_instance(heliocentric theory,model) inferred_instance(heliocentric theory,theory) inferred_instance(high middle ages,historical period) inferred_instance(high middle ages,middle ages) inferred_instance(high middle ages,time) inferred_instance(history,field of study) inferred_instance(history,humanities) inferred_instance(inductive method,classical method) inferred_instance(inductive method,method) inferred_instance(industrial revolution,historical period) inferred_instance(industrial revolution,modern) inferred_instance(industrial revolution,time) inferred_instance(kepler moon theory,model) inferred_instance(kepler moon theory,theory) inferred_instance(late middle ages,historical period) inferred_instance(late middle ages,middle ages) inferred_instance(late middle ages,time) inferred_instance(magnetism,phenomenon) inferred_instance(mathematics,field of study) inferred_instance(mathematics,science) inferred_instance(medicine,field of study) inferred_instance(medicine,science) inferred_instance(microscope,event) inferred_instance(microscope,invention) inferred_instance(moon phases,phenomenon) inferred_instance(netwon law of inertia,law) inferred_instance(netwon law of inertia,model) inferred_instance(newton law of acceleration,law) inferred_instance(newton law of acceleration,model) inferred_instance(newton law of reciprocal actions,law) inferred_instance(newton law of reciprocal actions,model) inferred_instance(newton universal gravitation,law) inferred_instance(newton universal gravitation,model) inferred_instance(nova,astro area) inferred_instance(nova,place) inferred_instance(nova,star) inferred_instance(philosophy,field of study) inferred_instance(philosophy,humanities) inferred_instance(physics,field of study) inferred_instance(physics,science) inferred_instance(planetary motion,phenomenon) inferred_instance(pre-socratic method,classical method) inferred_instance(pre-socratic method,method) inferred_instance(ptolemaic theory,model) inferred_instance(ptolemaic theory,theory) inferred_instance(religion,field of study) inferred_instance(religion,humanities) inferred_instance(scientific revolution,early modern) inferred_instance(scientific revolution,historical period) inferred_instance(scientific revolution,time) inferred_instance(sillogism,deductive argument) inferred_instance(sillogism,mode of reasoning) Appendix A Expanded queries 106 inferred_instance(telescope,event) inferred_instance(telescope,invention) inferred_instance(thermometer,event) inferred_instance(thermometer,invention) inferred_instance(thermoscope,event) inferred_instance(thermoscope,invention) inferred_instance(tides,phenomenon) inferred_instance(university of Pisa,academic organization) inferred_instance(university of Pisa,university) show inferred instances of(place). ?- show_inferred_instances_of(place). Egypt England Frombork Germany Graz Italy Kingdom of Poland Knudstrop Nicaea Pisa Prague Royal Prussia Scania Warmia nova show indirect instances predicate ?- show_indirect_instances. indirect_instance(Almagest,document) indirect_instance(Archimedes,person) indirect_instance(Archimedes,philosopher) indirect_instance(Archimedes,scientist) indirect_instance(Aristotle,person) indirect_instance(Armonia Nova,document) indirect_instance(Bacon,person) indirect_instance(Bacon,philosopher) indirect_instance(Brahe,person) indirect_instance(Brahe,philosopher) indirect_instance(Brahe,scientist) indirect_instance(Consequences of the Observations of Capillarity Phenomena,docu ment) indirect_instance(Copernicus,person) indirect_instance(Copernicus,scientist) indirect_instance(Darwin,person) indirect_instance(Darwin,scientist) indirect_instance(De Generatione,document) indirect_instance(De Humani Corporis Fabrica,document) indirect_instance(De Revolutionibus,document) indirect_instance(De lamark,person) indirect_instance(De lamark,scientist) Appendix A Expanded queries 107 indirect_instance(Democritus,person) indirect_instance(Democritus,philosopher) indirect_instance(Descartes,person) indirect_instance(Descartes,scientist) indirect_instance(Dialogue Concerning the Two Chief World System,document) indirect_instance(Egypt,geopolitical area) indirect_instance(Egypt,place) indirect_instance(Einstein,person) indirect_instance(Einstein,scientist) indirect_instance(Elements,document) indirect_instance(England,geopolitical area) indirect_instance(England,place) indirect_instance(Euclid,person) indirect_instance(Euclid,philosopher) indirect_instance(Euclid,scientist) indirect_instance(Frombork,geopolitical area) indirect_instance(Frombork,place) indirect_instance(Galenus,person) indirect_instance(Galenus,scientist) indirect_instance(Galileo,person) indirect_instance(Galileo,scientist) indirect_instance(Germany,geopolitical area) indirect_instance(Germany,place) indirect_instance(Gilbert,person) indirect_instance(Gilbert,philosopher) indirect_instance(Graz,geopolitical area) indirect_instance(Graz,place) indirect_instance(Halley,person) indirect_instance(Halley,scientist) indirect_instance(Harmonice Mundi,document) indirect_instance(Harvey,person) indirect_instance(Harvey,scientist) indirect_instance(Hipparcus,person) indirect_instance(Hipparcus,scientist) indirect_instance(Hobbes,person) indirect_instance(Huygens,person) indirect_instance(Huygens,scientist) indirect_instance(Italy,geopolitical area) indirect_instance(Italy,place) indirect_instance(Kepler,person) indirect_instance(Kepler,scientist) indirect_instance(Kingdom of Poland,geopolitical area) indirect_instance(Kingdom of Poland,place) indirect_instance(Knudstrop,geopolitical area) indirect_instance(Knudstrop,place) indirect_instance(Leibniz,person) indirect_instance(Methaphysics,document) indirect_instance(Newton,person) indirect_instance(Newton,scientist) indirect_instance(Nicaea,geopolitical area) indirect_instance(Nicaea,place) indirect_instance(On the Magnet and Magnetic Bodies,document) indirect_instance(Pascal,person) indirect_instance(Philosophiae Naturalis Principia Mathematica,document) indirect_instance(Pisa,geopolitical area) Appendix A Expanded queries 108 indirect_instance(Pisa,place) indirect_instance(Prague,geopolitical area) indirect_instance(Prague,place) indirect_instance(Ptolemy,person) indirect_instance(Ptolemy,philosopher) indirect_instance(Ptolemy,scientist) indirect_instance(Roger Bacon,person) indirect_instance(Roman Empire period,historical period) indirect_instance(Roman Empire period,time) indirect_instance(Royal Prussia,geopolitical area) indirect_instance(Royal Prussia,place) indirect_instance(Scania,geopolitical area) indirect_instance(Scania,place) indirect_instance(Scheiner,person) indirect_instance(Scheiner,scientist) indirect_instance(Sidereus Nuncius,document) indirect_instance(University of Graz,academic organization) indirect_instance(Vesalius,person) indirect_instance(Vesalius,scientist) indirect_instance(Warmia,geopolitical area) indirect_instance(Warmia,place) indirect_instance(aristotelian method,method) indirect_instance(arts,field of study) indirect_instance(astrology,field of study) indirect_instance(astronomy,field of study) indirect_instance(biology,field of study) indirect_instance(chemistry,field of study) indirect_instance(copernicus theory,model) indirect_instance(crystalline spheres,belief) indirect_instance(crystalline spheres,model) indirect_instance(deductive method,method) indirect_instance(early middle ages,historical period) indirect_instance(early middle ages,time) indirect_instance(electron discovery,event) indirect_instance(evolutionary theory,model) indirect_instance(galilean theory,model) indirect_instance(galilean theory of tides,model) indirect_instance(generalization,mode of reasoning) indirect_instance(geocentric theory,model) indirect_instance(heliocentric theory,model) indirect_instance(high middle ages,historical period) indirect_instance(high middle ages,time) indirect_instance(history,field of study) indirect_instance(inductive method,method) indirect_instance(industrial revolution,historical period) indirect_instance(industrial revolution,time) indirect_instance(kepler moon theory,model) indirect_instance(late middle ages,historical period) indirect_instance(late middle ages,time) indirect_instance(mathematics,field of study) indirect_instance(medicine,field of study) indirect_instance(microscope,event) indirect_instance(netwon law of inertia,model) indirect_instance(newton law of acceleration,model) indirect_instance(newton law of reciprocal actions,model) Appendix A Expanded queries 109 indirect_instance(newton universal gravitation,model) indirect_instance(nova,astro area) indirect_instance(nova,place) indirect_instance(philosophy,field of study) indirect_instance(physics,field of study) indirect_instance(pre-socratic method,method) indirect_instance(ptolemaic theory,model) indirect_instance(religion,field of study) indirect_instance(scientific revolution,historical period) indirect_instance(scientific revolution,time) indirect_instance(sillogism,mode of reasoning) indirect_instance(telescope,event) indirect_instance(thermometer,event) indirect_instance(thermoscope,event) indirect_instance(university of Pisa,academic organization) A.1.2 Relation Mode queries: additional examples show inferred transitive predicate ?- show_inferred_transitive. [contain,Austria,Bohemia] [contain,Denmark,Knudstrop] [contain,Egypt,Alexandria] [contain,Europe,Denmark] [contain,Europe,Knudstrop] [contain,Europe,Prague] [contain,Germany,Graz] [contain,Italy,Pisa] [contain,Italy,Roma] [contain,galaxy,star] [extend,brahe theory,ptolemy theory] [extend,copernican theory,brahe theory] [extend,copernican theory,ptolemy theory] [extend,galilean theory,brahe theory] [extend,galilean theory,copernican theory] [extend,galilean theory,kepler theory] [extend,galilean theory,ptolemy theory] [influence,Copernicus,Brahe] [influence,Copernicus,Galileo] [influence,Copernicus,Kepler] [influence,Hipparcus,Ptolemy] [influence,Kepler,Brahe] show inferred inverse predicate ?- show_inferred_inverse. [contain,Alexandria,Egypt] [contain,Bohemia,Austria] [contain,Denmark,Europe] [contain,Graz,Germany] [contain,Knudstrop,Denmark] [contain,Pisa,Italy] Appendix A Expanded queries 110 [contain,Prague,Europe] [contain,Roma,Italy] [contain,star,galaxy] [explain,tides,galilean theory of tides] [explain,tides,kepler moon theory] [explain,tides,newton universal gravitation] [extend,brahe theory,copernican theory] [extend,copernican theory,galilean theory] [extend,kepler theory,galilean theory] [extend,ptolemy theory,brahe theory] [formulate,Aristotle motion theory,Aristotle] [formulate,geocentrism,Ptolemy] [influence,Brahe,Kepler] [influence,Galileo,Copernicus] [influence,Kepler,Copernicus] [influence,Ptolemy,Hipparcus] [invent,telescope,Galileo] [invent,thermometer,Galileo] [investigate,sunspot,Galileo] [is contained by,Austria,Bohemia] [is contained by,Denmark,Knudstrop] [is contained by,Egypt,Alexandria] [is contained by,Europe,Denmark] [is contained by,Europe,Prague] [is contained by,Germany,Graz] [is contained by,Italy,Pisa] [is contained by,Italy,Roma] [is contained by,galaxy,star] [is explained by,galilean theory of tides,tides] [is explained by,kepler moon theory,tides] [is explained by,newton universal gravitation,tides] [is extended by,brahe theory,ptolemy theory] [is extended by,copernican theory,brahe theory] [is extended by,galilean theory,copernican theory] [is extended by,galilean theory,kepler theory] [is formulated by,Aristotle,Aristotle motion theory] [is formulated by,Ptolemy,geocentrism] [is influenced by,Copernicus,Galileo] [is influenced by,Copernicus,Kepler] [is influenced by,Hipparcus,Ptolemy] [is influenced by,Kepler,Brahe] [is invented by,Galileo,telescope] [is invented by,Galileo,thermometer] [is investigated by,Galileo,sunspot] [is observed by,Bevis,supernova] [is observed by,Brahe,supernova] [is observed by,Galileo,sunspot] [is observed by,Scheiner,sunspot] [is published by,Brahe,De nova stella] [is published by,Copernicus,De revolutionibus orbium coelestium] [is published by,Kepler,Astronomia Nova] [is published by,Kepler,Harmonice Mundi] [is read by,Galileo,Elements] [is rejected by,Brahe,crystalline sphere hypothesis] [is written by,Aristotle,Methaphysics] Appendix A Expanded queries 111 [is written by,Copernicus,De Revolutionibus orbium coelestium] [is written by,Einstein,Consequences of the Observations of Capillarity Phenomen a] [is written by,Euclid,Elements] [is written by,Galileo,Dialogue Concerning the Two Chief World Systems] [is written by,Galileo,Dialogue Concerning the Two Chief World systems] [is written by,Galileo,Discourse on the Two Sciences] [is written by,Galileo,Sidereus Nuncius] [is written by,Galileo,The Assayer] [is written by,Gilbert,On the Magnet and Magnetic Bodies] [is written by,Harvey,De Generatione] [is written by,Newton,Philosophiae Naturalis Principia Mathematica] [is written by,Ptolemy,Almagest] [is written by,Vesalius,De Humani Corporis Fabrica] [observe,sunspot,Galileo] [observe,sunspot,Scheiner] [observe,supernova,Bevis] [observe,supernova,Brahe] [publish,Astronomia Nova,Kepler] [publish,De nova stella,Brahe] [publish,De revolutionibus orbium coelestium,Copernicus] [publish,Harmonice Mundi,Kepler] [read,Elements,Galileo] [reject,crystalline sphere hypothesis,Brahe] [write,Almagest,Ptolemy] [write,Consequences of the Observations of Capillarity Phenomena,Einstein] [write,De Generatione,Harvey] [write,De Humani Corporis Fabrica,Vesalius] [write,De Revolutionibus orbium coelestium,Copernicus] [write,Dialogue Concerning the Two Chief World Systems,Galileo] [write,Dialogue Concerning the Two Chief World systems,Galileo] [write,Discourse on the Two Sciences,Galileo] [write,Elements,Euclid] [write,Methaphysics,Aristotle] [write,On the Magnet and Magnetic Bodies,Gilbert] [write,Philosophiae Naturalis Principia Mathematica,Newton] [write,Sidereus Nuncius,Galileo] [write,The Assayer,Galileo] Appendix B Ontology statistics and glossary entries In this appendix we provide the overall statistics on the size of the ontology and the full description of the concepts in terms of glossary entries (upper and lower classes). B.1 Statistics of ontology size In this section we show the overall statistics of the History of Science ontology with regard to upper and lower classes, upper and instantiated relations and instances (Table B1). Moreover, Table B2 separately illustrates the number of all transitive, inverse and symmetrical relations. Upper classes 15 Lower classes 98 Upper relations 87 Instantiated relations 41 event relations 32 instances 121 Table B.1: Overview of the ontology size Upper relations 48 Upper inverse relations 98 Upper symmetric relations 87 Upper transitive relations 41 Table B.2: Statistics on the ontology size: transitive, inverse and symmetrical upper relations B.2 Glossary entries: upper classes Table B.3 illustrates the full description for all upper classes. 112 Appendix B Name Person Role Ontology statistics and glossary entries Type Concept Concept 113 Source VICODI ontology Role ontology/Wikipedia Synonym Individual Social Role Place Concept Location Phenomenon Concept – Field of study Concept Belief Concept Subject Area (from Wordnet) Credence Document Concept – Mode of reasoning Event Doctrine Concept Concept Concept Form of Reasoning – Ism (from Wordnet) Method Concept – Model Concept – Group of People Concept Social Group Natural Language Description A named individual that really existed or exists A set of connected behaviors, rights and obligations as conceptualised by actors in a social situation. It is mostly deﬁned as an expected behavior in a given individual social status and social position A topographic point located with respect to surface features of some region Any state or process known through the sense rather than by intuition or reasoning An academic or applied discipline which recognized experts and with core of accepted theory or practice Belief is usually deﬁned as a conviction to the truth of a proposition without its veriﬁcation, therefore it is a subjective mental interpretation of the perception results, own contemplation/reasoning or communication A bounded physical representation of body of information designed with the capacity (and usually intent) to communicate. Any form of reasoning used to support or justify conclusions Anything that happens at a given place and time A generic standpoint, a conception or view about a problem area or a problem. It is not as formalized and systematic as a theory, and its contents are limited to be thesis A way of doing something, especially a systematic way which implies an orderly logical arrangement A theoretical construct that represents something, with a set of variables and a set of logical and quantitative relationships between them People sharing a common identity in form of social and cultural background, interests, values. Wordnet Wordnet/SUMO Wordnet/SUMO Wikipedia Wikipedia Wikipedia (partially adapted) Wikipedia Michele Pasin’s ontology Wordnet/SUMO Wikipedia Time Concept Temporal Thing The class which subsumes all temporal concepts VICODI ontology and Wikipedia (partially adapted) Cyc ontology/AKT support ontology Table B.3: Glossary entities for upper classes B.3 Glossary entries: lower classes In this section we illustrate the glossary entries for all lower classes. Each upper class might have more than one level of lower classes. For instance the upper class Person has two lower level classes (Scientist and Philosopher ) which have a number of subclasses, respectively. Therefore, we make use of the term ‘lower classes’ to indicate all direct (ﬁrst level) classes of a upper concept and ‘subclasses’ to refer to all direct subclass of a ﬁrst level classes. B.3.1 Person: lower classes Type Concept Table B.4 illustrates all direct classes of Person. Name Scientist Synonym – Philosopher Concept – Natural Language Description A person who applies advanced knowledge of one of more sciences A specialist in philosophy Source Role Ontology Wordnet Table B.4: Person: lower classes Appendix B B.3.1.1 Ontology statistics and glossary entries 114 Scientist: subclasses Table B.5 illustrates all subclasses of Scientist, a ﬁrst level class of Person. Name Inventor Type Concept Synonym – Discoverer Astronomer Mathematician Concept Concept Concept – – – Naturalist Concept Chemist Concept Physician Concept Natural Historian Alchemist (archaic) Surgeon Natural Language Description A person who produces something for the ﬁrst time through the use of the imagination or of ingenious thinking and experiment A person who is the ﬁrst to observe something. A physicist who studies astronomy A person whose primary area of study and research is the ﬁeld of mathematics A biologist knowledgeable about natural history Source Merriam-Webster SUMO/Wordnet Wordnet Wikipedia Wordnet A person trained in the science of chemistry Wikipedia/Wordnet Anatomist Concept – A person who practices some type of human biological medicine An expert in anatomy Wikipedia Physicist Concept – A person who studies or practices physics Wordnet/MerriamWebster Wikipedia Table B.5: Scientist: subclasses B.3.1.2 Philosopher: subclasses Table B.6 illustrates all subclasses of Philosopher, a ﬁrst level class of Person. Name Natural pher Philoso- Type Concept Synonym Philosopher Nature of Metaphysician Concept – Epistemologist Concept – Logician Concept – Astrologer Concept Astrologist Cosmologist Concept – Natural Language Description A person who was involved in the study of nature and the physical universe. This word becomes obsolete after the development of modern science and when the term scientist was coined (1833) A person who is involved in the study of the branch of philosophy concerned with explaining the ultimate nature of reality, being, and the world, so called Metaphysics. A person who is involved in the study of Epistemology, the branch of Philosophy that studies the nature, methods, limitations, and validity of knowledge and belief A person who studies Logics, a science that deals with the principles and criteria of validity of inference and demonstration A person who practices Astrology, a group of systems, traditions, and beliefs in which knowledge of the relative positions of celestial bodies and related details is held to be useful in understanding, interpreting, and organizing information about personality, human affairs, and other terrestrial matters A person who studies Cosmology, a branch of Metaphysics that deals with the nature of the universe1 Source Wikipedia (adapted) Wikipedia (adapted) Wikipedia(adapted) Wikipedia/MerrianWebster(adapted) Wikipedia/MerriamWebster(adapted) Wikipedia/MerriamWebster Table B.6: Philosopher: subclasses Appendix B Ontology statistics and glossary entries 115 B.3.2 Role: lower classes Type Concept Table B.7 illustrates all direct classes of Role. Name Character Role Synonym – Occupational Role Concept Function Natural Language Description Character role represents a role performed by an actor as a set of actors and pertains to general categories of a person or a group of people in a particular social setting The actions and activities assigned to, required or expected of a person or group Source Role Ontology Wordnet/SUMO Table B.7: Role: lower classes B.3.3 Place: lower classes Table B.8 illustrates all direct classes of Place. Name Geographical Area Geopolitical Area Role Type Concept Synonym – Concept – Astro Area Concept Astronomical body/AstroObject (from Science ontology) Natural Language Description A geographic location, generally having definite boundaries A geographical area which is associated with some sort of political structure. This class includes Lands, Cities, districts of cities, counties, etc. The Class of all astronomical objects of signiﬁcant size. It includes planets, stars, and asteroids, as well as nebulae, galaxies, and constellations. Source SUMO/Michele ontology SUMO/Michele ontology Pasin’s Pasin’s SUMO/Science ontology Table B.8: Place: lower classes B.3.3.1 Geopolitical Area: subclasses Table B.9 illustrates all subclasses of Geographical Area, a ﬁrst level class of Places. Name City Type Concept Synonym – Province Concept – State Concept – Kingdom Concept Kingship Empire Concept Imperial sovereignty Natural Language Description A large and densely populated urban area; may include several independent administrative districts The territory occupied by one of the constituent administrative districts of a nation A politically organized body of people under a single government A politically organized community or major territorial unit having a monarchical form of government headed by a king or queen A major political unit having a territory of great extent or a number of territories or peoples under a single sovereign authority Source SUMO/Wordnet SUMO/Wordnet SUMO/Wordnet Merriam-Webster Merriam-Webster Table B.9: Geopolitical Area: subclasses Appendix B B.3.3.2 Ontology statistics and glossary entries 116 Astro Area: subclasses Table B.10 illustrates all subclasses of Astro Area, a ﬁrst level class of Place. Name Planet Type Concept Synonym – Star Concept – Galaxy Concept Extragalactic nebula Natural Language Description Any celestial body (other than comets or satellites) that revolves around a star A celestial body of hot gases that radiates energy derived from thermonuclear reactions in the interior A collection of star systems; any of the billions of systems each having many stars and nebulae and dust Source Science ontology/SUMO Wordnet/Science ontology SUMO/Science ontology Table B.10: Astro area: subclasses B.3.4 Name Humanities Field of study: lower classes Type Concept Table B.11 illustrates all direct classes of Field of study. Synonym Human Sciences Science Concept – Pseudo science Concept Quasi-science Natural Language Description The humanities are those academic disciplines which study the human condition using methods that are largely analytic, critical, or speculative, as distinguished from the mainly empirical approaches of the natural and social sciences Science is a system of acquiring knowledge based on the scientiﬁc method, as well as to the organized body of knowledge gained through such research. Any body of knowledge, methodology, belief, or practice that claims to be scientiﬁc or is made to appear scientiﬁc, but does not adhere to the basic requirements of the scientiﬁc method Source Wikipedia Wikipedia Wikipedia Table B.11: Field of study: lower classes B.3.5 Belief: lower classes Type Concept Table B.12 illustrates all direct classes of Belief. Name Philosophical belief Psychological belief Synonym – Concept – Religious belief Concept – Natural Language Description A sentence-like constructs implying a commitment to something, which involves intellectual assent A psychological belief is are represented in the mind as sentence-like constructs, the simplest form of mental representation which implies the existence of mental states and intentionality A religious belief usually relates to the existence, nature and worship of a deity or deities and divine involvement in the universe and human life Source Wikipedia(adapted) Wikipedia(adapted) Wikipedia Table B.12: Belief: lower classes Appendix B Ontology statistics and glossary entries 117 B.3.6 Name Book Document: lower classes Type Concept Table B.13 illustrates all direct classes of Document. Synonym – In book Concept Article Concept Journal Concept Manual Web page Concept Concept Thesis Concept Technical report Proceeding Concept Concept Natural Language Description A written work or composition that has been published (printed on pages bound together) A part of a book, which may be a chapter (or section – or whatever) and/or a range of pages Nonﬁctional prose forming an independent part of a – publication from journal or magazine A journal is a periodical dedicated to a particular sub– ject). Technical documentation A document connected to the World Wide Web and viewable by anyone connected to the internet who has a web browser Dissertation A treatise advancing a new point of view resulting from research; usually a requirement for an advanced academic degree A report published by a school or other institution, – usually numbered within a series An ofﬁcial record of a conference – Source Wordnet (Bibtex entry)/SUMO Wikipedia (Bibtex entry) SUMO/Wikipedia (Bibtex entry) Wordnet Wikipedia (Bibtex entry) Wordnet/SUMO Wordnet/SUMO Wikipedia (Bibtex entry) In proceedings Manuscript Concept – An article in a conference proceedings. MerriamWebster/Wikipedia(Bibtex entry) Wikipedia(Bibtex entry) Concept – A written or typewritten composition or document as distinguished from a printed MerriamWebster/Wikipedia Table B.13: Document: lower classes B.3.7 Mode of reasoning class: lower classes Type Concept Table B.14 illustrates all direct classes of Mode of reasoning. Name Abductive argument Deductive argument Synonym – Concept – Inductive ment argu- Concept – Analogical argument Concept – Natural Language Description A reasoning process that starts from a set of facts and derives their most likely explanation An argument in which the conclusion is necessitated by, or reached from, previously known facts. If the premises are true, the conclusion must be true An argument in which the premises of an argument are believed to support the conclusion but do not ensure it. It is used to ascribe properties or relations to types based on tokens (i.e., on one or a small number of observations or experiences); or to formulate laws based on limited observations of recurring phenomenal pattern An argument which consists in transferring from a particular subject to another Source Wikipedia (adapted) Wikipedia Wikipedia Wikipedia Table B.14: Reasoning: lower classes Appendix B Ontology statistics and glossary entries 118 B.3.8 Doctrine: lower classes Type Concept Table B.15 illustrates all direct classes of Doctrine. Name Scientiﬁc doctrine Philosophical doctrine Synonym – Concept – Natural Language Description Body of principles or teachings in the ﬁeld of science taught and accepted by a particular group. A generic standpoint or conception in the ﬁeld of philosophy. Philosophical doctrines often originates from a speciﬁc theory or philosophical system, but they exist as more abstract entities. Examples are ‘rationalism’, or ‘solipsism’: they do not refer to a theory in particular, but just indicate some generic tendency in looking at things, or a generic belief. Therefore they are only described as a set of ‘thesis’. Conceptions become in this way powerful categorizers of other theories. A doctrine such as ‘solipsism’ can be in fact treated in epistemology, or in moral theory. Source Wikipedia/Cambridge dictionary Michele Pasin’s ontology Table B.15: Doctrine: lower classes B.3.9 Event: lower classes Table B.16 illustrates all direct classes of Event. Name Discovery Type Concept Synonym – Natural Language Description The act of process of discovering something Invention Concept – The act of process of inventing something Experiment Concept – Observation Concept – An operation or procedure carried out under controlled conditions in order to discover an unknown effect or law, to test or establish a hypothesis, or to illustrate a known law The act of recognizing and noting a fact or occurrence often involving measurement with instruments Source Science ontology/MerriamWebster/Wordnet Science ontology/MerriamWebster/Wordnet Merriam-Webster MerriamWebster/Wordnet/SUMO Table B.16: Event: lower classes B.3.9.1 Observation: subclasses Table B.17 shows all subclasses of Observation, a ﬁrst level class of Event. Name Anomaly Type Concept Synonym – Natural Language Description (1) Anomaly is something which does not ﬁt into an accepted classiﬁcatory scheme or theory (general definition) (2) The position of a planet as deﬁned by its angular distance from its perihelion . Source (1) Dictionary of History of Science [10] (2) Wordnet/SUMO Table B.17: Observation: subclasses Appendix B Ontology statistics and glossary entries 119 B.3.10 Name Realism Philosophical doctrine class: subclasses Type Concept Table B.18 illustrates all subclasses of Philosophical doctrine, a ﬁrst level class of Doctrine. Synonym – Rationalism Concept – Idealism Concept – Natural Language Description (1)The philosophical doctrine that physical object continue to exist when not perceived (2) the philosophical doctrine that abstract concepts exist independent of their names The doctrine that knowledge is acquired by reason without resort to experience The philosophical theory that ideas are the only reality Source Michele Pasin’s ontology/Wordnet/SUMO Pragmatism Concept – Phenomenology Concept Dualism Concept Monism Concept Analytic Concept Intuitionism Concept Materialism Concept The doctrine that practical consequences are the criteria of knowledge and meaning and value A philosophical doctrine proposed by Edmund Husserl – based on the study of human experience in which considerations of objective reality are not taken into account The doctrine that reality consists of two basic opposing – elements, often taken to be mind and matter (or mind and body), or good and evil The doctrine that reality consists of a single basic sub– stance or element Philosophical doctrine referring to the British and Ameri– can philosophers of the twentieth century, focused on language and its logical aspects The doctrine that knowledge is acquired primarily by in– tuition Physicalism The philosophical doctrine2 that matter is the only reality Michele Pasin’s ontology/Wordnet/SUMO Michele Pasin’s ontology/Wordnet/SUMO Michele Pasin’s ontology/Wordnet/SUMO Michele Pasin’s ontology/Wordnet/SUMO Wordnet/SUMO Wordnet/SUMO Michele Pasin’s ontology Mecanicism Concept – Mentalism Concept – The philosophical doctrine that all phenomena can be explained in terms of physical or biological causes The doctrine that mind is the true reality and that objects exist only as aspects of the mind’s awareness Michele Pasin’s ontology/Wordnet/SUMO Michele Pasin’s ontology/Wordnet/SUMO Michele Pasin’s ontology/Wordnet/SUMO Michele Pasin’s ontology/Wordnet/SUMO Table B.18: Philosophical doctrine: subclasses B.3.11 Method: lower classes Type Concept Table B.19 illustrates all direct classes of Method. Name Classical method Synonym – Scientiﬁc method Concept – Natural Language Description The classical method in scientiﬁc enquiry derives from Aristotle, who distinguished the forms of approximate and exact reasoning, set out the threefold scheme of abductive, deductive, and inductive inference, and also treated the compound forms such as reasoning by analog Any method which considers observations and experiments as systematical way of discovering knowledge or science. Source Wikipedia The Oxford Companion to the History of Modern Science [47] Table B.19: Method: lower classes Appendix B Ontology statistics and glossary entries 120 B.3.12 Name Hypothesis Model: lower classes Type Concept Table B.20 illustrates all direct classes of Model. Synonym – Theory Concept – Law Concept Law of science Natural Language Description (1) An unveriﬁed proposal intended to explain certain facts or observations (2) A hypothesis consists either of a suggested explanation for a phenomenon or of a reasoned proposal suggesting a possible correlation between multiple phenomena A systemic conceptual construction, with a coherent and organic architecture. It explains a speciﬁc phenomenon (or a set of phenomena) and answers an existing problem. It comprises concepts, statements and can be related to at an argument (1) A generalization that describes recurring facts or events in nature. (2) Laws differ from hypotheses, theories, postulates, principles, etc., in that they are analytic statements, usually with an empirically determined constant Source (1) Wikipedia (2) Wordnet/SUMO Michele Pasin’s ontology (1) Wordnet/SUMO (2) Wikipedia Table B.20: Model: lower classes B.3.13 Group of people: lower classes Type Concept Table B.21 illustrates all direct classes of Group of people. Name Academic organization Non academic organization Synonym – Concept Cultural organization Philosophical school Concept – Scientiﬁc school Concept – Natural Language Description Any formal institution or group involved in education and research All organizations related to culture including religion which do not belong to any academic institutions or department Particular schools of thought, styles of philosophy, or descriptions of philosophical ideas attributed to a particular group or culture Particular schools of thought, corpus of scientiﬁc doctrines, or descriptions of scientiﬁc ideas attributed to a particular group or culture Source VICODI ontology VICODI (adapted) ontology Wikipedia(adapted) Wikipedia(adapted) Table B.21: Group of people: lower classes B.3.13.1 Academic organization: subclasses Table B.22 shows all subclasses of Academic organization, a ﬁrst level class of Group of people. Name University Type Concept Synonym – Natural Language Description Any institution of higher education Source VICODI ontology Table B.22: Academic organization: subclasses Appendix B Ontology statistics and glossary entries 121 B.3.14 Name Time point Time: lower classes Type Concept Table B.23 illustrates all direct classes of Time. Synonym Instant Time interval Concept – Time unit Concept Time measure Natural Language Description Time point is the class of all extensionless points on the universal timeline and is expressed as number consisting of year-month-day Time interval is the class of all sets of all points in the time line Time unit is the class of all unit of measurement used to quantify time (minutes, hours, years, etc) Source Michele Pasin’s ontology/AKT support ontology(adapted) KSL ontology Historical period Concept Human time period Historical period is the class of all named time periods which are categorized into discrete blocks AKT support ontology/OWL/KSL time ontology Wikipedia (Periodization entry) Table B.23: Time: lower classes B.3.14.1 Historical period:subclasses Table B.24 shows all subclasses of Historical period, a ﬁrst level class of Time. Name Ancient Type Concept Synonym – Middle Ages Concept Medieval period Early modern Concept – Modern Concept Modern times Natural Language Description Ancient history is the study of the written past from the beginning of human history until the Early Middle Ages. Although the ending date of ancient history is disputed, currently most Western scholars use the fall of the Western Roman Empire in AD 476 as the end of ancient European history The Middle Ages are a period commonly dated from the 5th century fall of the Western Roman Empire until the fall of the Eastern Roman Empire in the 15th century The early modern period is a term used by historians to refer to the period in Western Europe and its ﬁrst colonies which spans the two centuries between the Middle Ages and the Industrial Revolution. The beginning of the early modern period is not clear-cut, but is generally accepted to be in the late 15th century or early 16th century. The end date of the early modern period is usually associated with the Industrial Revolution, which began in Britain in about 1750 The term Modern Times is used by historians to loosely describe the period of time immediately following what is known as the Early Modern. This historical period spans from Enlightenment until today Source Echo portal/Wikipedia Echo portal/Wikipedia Echo portal/Wikipedia Echo portal/Wikipedia Table B.24: Historical period: subclasses

Ontology-based representation and reasoning about the history of science more

Log In

Reset Password