Ontologies For What? Ontologies. Origin and History (II) Origin and History (I)

Transcription

1 Ontologies For What? Ontologies Lack of a shared understanding leads to poor communication => People, organizations and software systems must communicate between and among themselves Disparate modeling paradigms, languages and software tools limit => Interoperability [Uschold, Gruninger, 96] => Knowledge sharing & reuse Steffen Staab - 1 Steffen Staab - 2 Origin and History (I) Ontology... a philosophical discipline, branch of philosophy that deals with the nature and the organisation of reality Science of Being (Aristotle, Metaphysics, IV, 1) Tries to answer the questions: Origin and History (II) Humans require words (or at least symbols) to communicate efficiently. The mapping of words to things is only indirect possible. We do it by creating concepts that refer to things. The relation between symbols and things has been described in the form of the meaning triangle: Concept What is being? What are the features common to all beings? Jaguar [Ogden, Richards, 1923] Steffen Staab - 3 Steffen Staab - 4

2 Origin and History (III) In recent years ontologies have become a hot topic of interest.... Human Agent 1 Human and machine communication (I) Human Agent 2 Machine Agent 1 [Maedche et al., 2002] Machine Agent 2 Here, an ontology refers to an engineering artifact: It is constituted by a specific vocabulary used to describe a certain reality, plus a set of explicit assumptions regarding the intended meaning of the vocabulary. Thus, ontologies describe a formal specification of a specific domain: Shared understanding of a domain of interest Formal and machine executeable model of a domain of interest HA1 exchange symbol, e.g. via nat. language JAGUAR Internal models HA2 commit commit Ontology Description Formal Semantics Ontology commit a specific domain, e.g. animals MA1 exchange symbol, e.g. via protocols commit Formal models MA2 Symbol Concept Things Meaning Triangle Steffen Staab - 5 Steffen Staab - 6 Human and machine communication (II) People can t share knowledge if they do not speak a common language Young Person Syntax is not enough for machine communication, e.g. B2B Order information: <Product> <type>car</type> <Name>Daimler 500 SLK </Name> <Price> $</Price> </Product> Engineer Designer Banker Steffen Staab - 7 Bestellinformation: <Auto> <Name>Daimler 500 SLK </Name> <Preis> </Preis> </Auto> Definition (I) When talking about an ontology it is important to have a common definition on what one is talking about. The problem of exactly defining what an ontology is hasn t been solved by the community. The relation between ontology and knowledge base is also a point of discussion. In the following we will present different existing natural language-based definitions, how knowledge bases relate to ontologies Steffen Staab - 8

3 Definition (II) An ontology is an explicit specification of a conceptualization [Gruber, 93] An ontology is a shared understanding of some domain of interest. [Uschold, Gruninger, 96] What else exists in literature? There are many definitions, an ontology (in our sense) is a sound semantic basis to define meaning general logical theory constituted by a vocabulary and set of statements in some logical language foundation for communication between human and machine agents Steffen Staab - 9 What is a concept? Different communities have different notions on what a concept means: Formal concept analysis (see They talk about formal concepts Description Logics (see They talk about concept labels ISO-704:2000 Terminology Work: (see Often the classical notion of a frame in AI or a class in OO modeling is seen as equivalent to a concept. Steffen Staab - 10 Ontology & Natural Language It is important to emphasize that there is a m:n relationship between words and concepts Example Ontology: C = {c 1,c 2,c 3 }, R = {r 1 }, H C (c 2,c 1 ), r 1 (c 2,c 3 ), Lexicon: L C = {person, employee, organisation}, L R ={worksat} F(person) = c 1, F(employee) = c 2, F(organisation) = c 3, This means practically: G(works at) = r 1... different words may refer to the same concept a word may refer to several concepts Ontologies languages should provide means for making this difference explicit. organisation works at person employee c3 r 1 (c 2,c 3 ), c1 H C (c 2,c 1 ) c Steffen Staab - 11 Steffen Staab - 12

4 Ontology vs. Knowledge/Data Bases There is no clear separation between ontology and knowledge base Types of Ontologies (I) [Guarino, 98] describe very general concepts like space, time, event, which are independent of a particular problem or domain. It seems reasonable to have unified top-level ontologies for large communities of users. Example: person medication Aspirin Ann cured-with Aspirin pill-1 pill-2 taken-aspirins taken-aspirins Often it remains a modeling decision if something is modeled as concept or as instance. In many applications meta-modeling means are required. describe the vocabulary related to a generic domain by specializing the concepts introduced in the top-level ontology. These are the most specific ontologies. Concepts in application ontologies often correspond to roles played by domain entities while performing a certain activity. describe the vocabulary related to a generic task or activity by specializing the top-level ontologies. Steffen Staab - 13 Steffen Staab - 14 Ontologies and their Relatives (I) There are many relatives around: Controlled vocabularies, thesauri and classification systems available in the WWW, see Classification Systems (e.g. UNSPSC, Library Science, etc.) Thesauri (e.g. Art & Architecture, Agrovoc, etc.) Lexical Semantic Nets WordNet, see EuroWordNet, see Topic Maps, (e.g. used within knowledge management applications) Ontologies and their Relatives (II) Catalog / ID Terms/ Glossary Thesauri Informal Is-a Formal Is-a Formal Instance Frames Value Restrictions General logical constraints Axioms Disjoint Inverse Relations,... In general it is difficult to find the border line! Steffen Staab - 15 Steffen Staab - 16

5 Ontologies - Some Examples General purpose ontologies: WordNet / EuroWordNet, The Upper Cyc Ontology, IEEE Standard Upper Ontology, Domain and application-specific ontologies: RDF Site Summary RSS, UMLS, KA2 / Science Ontology, RETSINA Calendering Agent, AIFB Web Page Ontology, Web-KB Ontology, 11/www/wwkb/ Dublin Core, Meta-Ontologies Semantic Translation, RDFT, Evolution Ontology, Ontologies in a wider sense Agrovoc, Art and Architecture, UNSPSC, DTD standardizations, e.g. HR-XML, Steffen Staab - 17 UMLS Unified Medical Language System Framework consisting of several knowledge bases and according tools Goal: Improvement of knowledge management in information systems with medical context Publisher: US National Library of Medicine Steffen Staab - 18 UMLS - Goals Automated processing of biomedical and health care related terms Support of application developers Integration of several kinds of information, e.g.: Scientific literature Medical records Epidemiological data UMLS Components Linked knowledge bases Metathesaurus Semantic Network SPECIALIST Lexicon & NLP system Tools UMLS Knowledge Source Server MetamorphoSys Lexical Tools Steffen Staab - 19 Steffen Staab - 20

6 UMLS Methathesaurus Data stock Terms related to medicine and health care Additional detailed information Integration of heterogeneous thesauri and classifications Goals Linkage between different terms/views of the same concept Generation of relations between different concepts Metathesaurus Data sources Alcohol and Other Drug Thesaurus Physicians' Current Procedural Terminology International Classification of Diseases (ICD) Gene Ontology (GO) Health Level Seven Vocabulary (HL7) MEDLINE Medical Subject Headings (MESH) Systematized Nomenclature of Medicine (SNOMED) More than 100 vocabularies more than 1 million concepts Steffen Staab - 21 Steffen Staab - 22 Metathesaurus Structure Metathesaurus Excerpt Elements with unique identifier Concepts (CUI) Strings: different languages/spellings, etc. (SUI) Atoms: every appearance of a string (AUI) Terms: Gruops lexically related strings (TUI) Relations between concepts Attributs for conceps, atoms und relations Concept (CUI) C Atrial Fibrillation (preferred) Atrial Fibrillations Auricular Fibrillation Auricular Fibrillations Terms (LUIs) L Atrial Fibrillation (preferred) Atrial Fibrillations Strings (SUIs) S Atrial Fibrillation (preferred) S (plural variant) Atrial Fibrillations Atoms (AUIs) A Atrial Fibrillation (from MSH) A Atrial Fibrillation (from PSY) A Atrial Fibrillations (from MSH) Contradictory representations are adopted No overall, consistent ontology L (synonym) Auricular Fibrillation Auricular Fibrillations S Auricular Fibrillation (preferred) S (plural variant) Auricular Fibrillations A Auricular Fibrillation (from PSY) A Auricular Fibrillations (from MSH) Steffen Staab - 23 Steffen Staab - 24

7 Metathesaurus Relations Source internal relations Within one source vocabulary Indicated e.g. by hierarchies, cross references Statistical Relations, e.g. joint appearance of concepts in articles, several diagnose codes for the same patient, etc. Source overlapping relations Mappings between multiple source vocabularies Inclusion of orphaned concepts into more detailed contexts of other vocabularies UMLS Semantic Network Goal: Consistent classification of concepts from the Metathesaurus Data stock: Topic categories: Semantic Types Relations between Semantic Types: Semantic Relations 135 Semantic Types, 54 Semantic Relations Each concept in the Metathesaurus is associated to at least one Semantic Type Steffen Staab - 25 Steffen Staab - 26 Semantic Network Structure Semantic Network Excerpt Main concepts for Semantic Types Organisms Anatomical structures Biologic function Chemicals Events Physical objects Concepts or ideas Main concepts for Semantic Relations is a physically related to spatially related to temporally related to functionally related to conceptually related to Steffen Staab - 27 Steffen Staab - 28

8 UMLS SPECIALIST SPECIALIST Lexicon English dictionary with focus on biomedical terms Base for the SPECIALIST NLP System Content: syntactic, morphological und orthographic information SPECIALIST NLP Tools Management of lexical variants and text analysis in the context of biomedicine Support of integration into information systems Gene Ontology Goal of the GO Project Consistent descriptions of genes/gene products spanning multiple databases Three ontologies for description of gene products from different perspectives (overlapping species) Biological process Cellular component Molecular function Result: uniform queries over the different databases are possible Steffen Staab - 29 Steffen Staab - 30 Gene Ontology Three-step approach Development of ontologies Linkage between ontologies Development of dedicated tools Structure of ontologies: Directed Acyclic Graphs GO Biological process bp: Series of events, by chaining molecular functions Examples Cellular physiological process Signal transmission Pyrimidine metabolism A gene product is used in one or multiple biological processes Steffen Staab - 31 Steffen Staab - 32

9 GO Cellular components Cc: should be obvious One gene product is related to one or several cellular components GO Molecular function mf: activities on molecular level, mostly catalytic or binding Examples Transport Adenylate cyclase Phosphate synthase activity One gene product has one or multiple molecular functions Steffen Staab - 33 Steffen Staab - 34 Cellular process Cellular physiological process GO Excerpt All Biological process Physiological process Metabolism Database search, e.g. Gene products with a specific annotation All annotations to one gene product Applications Cellular metabolism Alcohol metabolism is a Steffen Staab - 35 Steffen Staab - 36

10 Applications GoFigure: Predicting functionality of gene sequences GO annotated DBs contain gene sequences BLAST algorithm: identification of annotated homologues to unknown gene sequences Applications Cancer research: Functional clustering of expressed genes in tumor cells (Microarray analysis) Steffen Staab - 37 Steffen Staab - 38 Galen Generalized Architecture for Languages, Encyclopaedias and Nomenclatures in Medcine Infrastructure for the integration of terminology in clinical information systems Concept system independent of natural language and used coding system Publisher: OpenGalen Non profit organisation (U. Manchester, U. Nijmegen) Galen Representation of formal definitions of concepts and their relations Move from simple enumeration to composition of concepts Separation of grammar and lexicon Realization: Terminology: GALEN Common Reference Model (CORE) Language: GALEN Concept Representation Language (GRAIL) Implementation: GALEN Terminology Server Steffen Staab - 39 Steffen Staab - 40

11 CORE Model High level ontology General concepts Common Reference Model Reusable concepts of domains such as anatomy, diseases, etc. Subspecialty extensions Extensions for specific applications and subdomains specific body regions, kinds of surgery, etc. Model of surgical procedures and others Composed concepts of the upper layers High level ontology (Upper parts of the hierarchy) GeneralisedStructure GeneralisedProcess GeneralisedSubstance Feature DomainCategory Aspect ModifierConcept State Selector Status Unit Modality Role GeneralLevelOfSpecification Collection Steffen Staab - 41 Steffen Staab - 42 ShuntStructure Common Reference Model (Excerpt) BodyConnection AnastomosisStructure Fistula BypassStructure BodyStructure ArbitraryBodyConstruct GenericBodyStructure SurfaceBodyPart SurfaceSubpart InternalBodyPart ExternalOrgan SurfaceOpening JunctionalBodyPart BodyPart InternalOrgan GITractBodyPart Steffen Staab - 43 NervousSystemPart SkeletalStructure Common Reference Model Extension by additional hierarchy axes by attributes, e.g. hastopology hasseverity hasshape HasDivision hassolidregion hasleftrightselector hasstructuralcomponent ismadeof Contains PassesThrough Example (HeartAtrium which hasleftrightselector rightselection) name RightHeartAtrium. (HeartVentricle which hasleftrightselector rightselection) name RightHeartVentricle. RightSideOfHeart sensiblyandnecessarily hasspecificstructuralcomponent [RightHeartVentricle, RightHeartAtrium]. Steffen Staab - 44

12 GRAIL Formal language for expressing the CORE Model Based on Description Logics and Conceptual Graphs Additional elements for Extension of the classification ( essential criteria, necessary statements ) Embedding of definitions Uniform classification of categories and individuals (no A-Box) Constraints on negation, disjunction, equality, quantification und referencing GALEN Terminology Server Access to the encapsulated CORE Model via several service interfaces Concept services: classification of concepts into a hierarchy Language services: transformation between concepts and natural language terms Coding services: discovery of codes in classification systems for a concept Indexing services: access to detailed information about a specific concept Steffen Staab - 45 Steffen Staab - 46 Sources UMLS Gene Ontology Galen Steffen Staab - 47