Developing Ontologies based on RDFOWL Semantic Web languages
Developing Ontologies based on RDF-OWL Semantic Web languages (for information sharing & knowledge representation) How-to@2 : 2006 -04 -13 David George
How do we share data, information & knowledge? Demonstrated in several dimensions: • we share amongst people – using HTML, dynamically with DBs • share between database systems – linking DBs: DBs • between organisations – exchanging data via XML and XSL transforms. • in searches – autonomous & collaborative intelligent software agents. but. . . sharing data requires understanding of the context of terms. . . “the semantics of data” using metadata. Hence Semantic Web would provide shared understanding using metadata vocabularies (using an ontological approach). 2
We have the Web: a Global Information Space Some current Web statistics • Approx. 40 m web sites? • Circa 10 -15 billion pages? (Google) Semantic Web share <http: //swoogle. umbc. edu/> • 0. 001% usable Semantic Web files • 0. 00006% are Ontologies 3
Result: effective query (precision) compromised Example: Query about Cook discovering New Zealand? Cook New Zealand 4
What is the Semantic Web? • A project aimed to make web pages machine understandable. “An extension of the current Web, … information given well-defined meaning, …enabling computers and people to work in co-operation” (Berners-Lee et al, 2001) • A universal medium for information exchange; where Ontologies are viewed as a pivotal component in giving meaning or semantics. • A solution based on “XML-based” RDF (Resource Description Framework) and OWL Ontology languages (W 3 C, 2004). • Expected that Semantic Web will have a role in Web Services and Grid Computing. 5
Descriptions of Ontology Socrates & Aristotle 400 -360 BC - philosophy of being: “Onto” Some definitions of Ontologies: • “An ontology is an explicit [formal] specification of a [shared] conceptualisation” (Gruber, 1993, [Borst, 1997]) • “A logical theory which gives an explicit, partial account of a conceptualisation” (Guarino & Giaretta, 1995) • “Conceptualisation refers to abstract model, . . formal refers to machine-readable, . . and shared reflects notion that ontology captures consensual knowledge shared by the group” (Studer et al, 1998) 6
Ontology Examples “Ontology” covers a range of things • Term lists - Catalogues for on-line shopping e. g. Amazon. • Dublin Core meta standards for the Web. • Linguistic structures – e. g. Thesauri like Word. Net. • Informal hierarchies or Taxonomies e. g. Yahoo & DMOZ directories. • Detailed formal classifications e. g. UNSPSC • Formal subsumption hierarchies like Gene Ontology. • OWL DL based ontologies • Domain-independent or philosophically inspired: Cyc, Sowa, IEEE SUMO Glossaries & Data Dictionaries Thesauri & Taxonomies Formal Ontologies & Inferencing 7
Why develop Ontologies? • Makes domain descriptions and assumptions explicit by defining: – Concepts – relationships and attributes of concepts – constraints on properties – Instances • Enables re-use of terms and relationships to avoid reinventing descriptions. • Allows domain knowledge to be separated from operational information. • Helps to manage the information explosion caused by the Web. 8
What do yet we have at present? Web! Well, we don’t a Semantic 9
But we do have HTML and XML! 10
HTML Document header table <h 3> para <table> Dept. of Computing <b> Subject <tr> <td> Name <tr> <td> David George Room CM 222 text Ontology <b> <link> emailto: dgeorge@ uclan. ac. uk • HTML syntax describes layout • Simply a presentation of content • Good for humans; not for machines 11
XML Document Object Model <person> <name> <subject> <locn> Ontology <lastname> David <firstname> <dept> <room> George Dept. of Computing CM 222 <email> mailto: dgeorge@ uclan. ac. uk Improve the description for • XML structures information not page. understanding? • Nested elements in tree hierarchy. • Uses syntax to differentiate data. • Universal standard for data exchange. • Good for machines (and humans). 12
How can Semantic Web languages improve our interpretation of information? 13
My Research • Using Semantic Web technologies to demonstrate that RDF-based language and Ontology can be used to integrate and share information. • Examining the way in which different Ontology structures can be developed and mapped together. • Motivating example will relate to Geographical (or Cosmological) domain – some early work. • Developed an interface to query an Ontology. • Some of the following slides relate to these domain concepts. 14
Geographic Ontology Layers rivers demographics economic Water Utility L. A. Planning pipelines settlements relief 15
Cosmological Ontologies 16
RDF Building Block 17
RDF (Resource Description Framework) • W 3 C standard (2004) for content (resource) description. • RDF is machine-processable; but not for humans, as we’ll see! • subject RDF parser interpretes common structures to convey semantics. predicate object • Built on subject, predicate, object triples [a statement] • A statement may say: <student> <lastname> is <George> • For example: • RDF uses the URI references like <http: //someurl>for describing s, p, o “resources” • Resources are anything that can be identified on the Web. 18
RDF Model • Previous RDF example represents a Directed Acyclic Graph (directed graph with no directed cycles v a tree) • statement triple (Subject, predicate, object) allows nodes to be linked across the Web, e. g. student URL and computing/semanticweb URL. http: //www. uclan. ac. uk/people/member 19
RDF nodes 20
RDF nodes n RDF is useful for describing data. n Basis for Ontology structures using OWL Web Ontology Language. n RDF graphs form complex directed graphs of linked triples, across the Web. 21
Semantics through more Metadata a a a Current Web a a a Semantic Web? (Kiryakov et al, 22 2004)
Semantic (Shadow) Web a a a 23
How do we define metadata? Vocabulary • Data/Information • Metadata Ontology used by described by specified by • Vocabularies • Semantic Web languages Terms Metadata formalised by described by Content Data 24
OWL (Web Ontology Language) RDF Schema layer rdfs: Resource rdfs: sub. Class. Of rdfs: Class rdf: Property rdfs: Class OWL Ontology layer rdf: type owl: Object. Property owl: Pop. Group rdfs: Domain rdfs: sub. Class. Of rdf: type owl: connected. To owl: City owl: Highway rdfs: Range rdfs: sub. Class. Of owl: Motorway Instance layer Manchester owl: connected. To M 62 25
Role of Ontology in a Semantic Web DB KB DB 26
Hierarchy of Ontologies Imprecise – Abstract - Generalised Upper-level: domain independent, general concept terms and relationships like space, time, matter, objects and events. Upper-level Ontology Generic domain concepts, e. g. medical, pharmaceutical, travel; Generic tasks like buying or selling. Domain-level Ontology Task-level Ontology Application-level Ontology specialisations of both domain and task, e. . g. flight travel by a specific travel organisation. Precise – Real - Specialised [Ontology classification (Guarino, 1998)] 27
“Upper-level” Ontologies (Chandrasekaran et al. , 1999) • Can represent the “starting points” for a field of study. • Required when working in large groups, i. e. generalisation is required to gain consensus on agreed terms 28
Mapping Ontology Levels Thing Object Concrete Physical Object Solar System Galaxy Systems Sun Planetary Exploration Domain Celestial Mechanics Cosmic Microwaves Planetary System Upper level Information Process Astromomy Nuclear Fusion Stellar Systems Abstract Information Object Physical Process Solar Physics Cosmology Process Planetary Characteristics Manned Exploration Application 29
Mapping Geographical Layers (1) 30
Mapping Geographical Layers (2) 31
• Potentially many translating functions • Complexity, scalability & maintenance • No consensus issues Ontology Mapping Ontology A Ontology B Ontology C Top-level Ontology • Resource ontologies are clustered on the basis of similarity. • General concepts are shared at a higher level. • Flexible and scalable Ontology A, B, C, D Ontology D One-to-one mapping Ontology A, B Ontology A Ontology B Ontology A, B, C, D Ontology A Ontology D Shared Ontology B Ontology C Ontology D Clustered Ontologies Potential consensus problems in agreeing a standard between many users 32
Importing Ontology Structures 33
OWL ontology imports <rdf: RDF xmlns: owl="http: //www. w 3. org/2002/07/owl#" xmlns=http: //www. owl-ontologies. com/unnamed. owl#> <owl: Ontology rdf: about=""> <owl: imports rdf: resource="http: //193. 61. 241. 101/union/british. owl"/> <owl: imports rdf: resource="http: //193. 61. 241. 101/union/american. owl"/> </owl: Ontology> <owl: Class rdf: ID="Retail. Operation"> <rdfs: sub. Class. Of> <owl: Class rdf: ID="Corporate. Entity"/> </rdfs: sub. Class. Of> </owl: Class> <owl: Class rdf: ID="Distribution. Operation"> <rdfs: sub. Class. Of rdf: resource="#Corporate. Entity"/> </owl: Class> </rdf: RDF> 34
Complexity in mapping Equivalence Different specifications of descriptions of relationships, when importing ontologies, will produce differing degrees of mappings, e. g. using equivalence, disjoint, and sub-class relations. This can superimpose additional complexity, for example recursive relations in equivalence. 35
OWL Web Ontology Language • Three species of OWL: – OWL Lite – class, object & property terms, inc. inverse, transitive, equivalence, difference. – OWL DL – greater expressivity. • inc. disjoint, min/max cardinality, union, complement, intersection • complex but computationally decideable. – OWL Full – most expressive but computationally problematic, e. g. answers not in finite time. • OWL based on “Open World Assumption” (OWA): (If not exists, will say NO only if can prove false). • DBs based on “Closed World Assumption” (CWA): (If not exists, will say NO). 36
Description Logic Expressions OWL Constructor Protégé-OWL Example intersection. Of C ⊓D Person union. Of C ⊔D Male ⊓ Employee ⊔ Female Meaning AND OR complement. Of ¬C ¬Male NOT one. Of {x y z} {Fiat BMW Ford} the set of some. Values. From ∃RC ∃ has. Vehicle Car SOME (from) all. Values. From ∀RC ∀ has. Vehicle Car ONLY (from) min. Cardinality R≥N has. Vehicle ≥ 3 MIN max. Cardinality R≤N has. Vehicle ≤ 3 MAX cardinality R=N has. Vehicle = 3 EXACTLY has. Value R∋I has. Vehicle ∋ Ford HAS (specific indiv. ) Ref: C, D = Class, I = Individual, R = Restriction 37
OWL RDF/XML-based Ontology Graph <owl: Class rdf: ID="Population. Group"/> <owl: Class rdf: about="#Town"> <rdfs: sub. Class. Of rdf: resource="#Population. Group"/> </owl: Class> <Town rdf: ID="Nelson"> <grid. Ref rdf: datatype="#string">2 E 52 N</grid. Ref> </Town> <owl: Class rdf: ID="City"> <rdfs: sub. Class. Of rdf: resource="#Population. Group"/> </owl: Class> <City rdf: ID="Liverpool"> <grid. Ref rdf: datatype=“#string">3 E 52 N</grid. Ref> </City> <owl: Datatype. Property rdf: ID="grid. Ref"> <rdfs: domain rdf: resource="#Population. Group"/> </owl: Datatype. Property> 38
Specifying Descriptions & Constraints 39
Ontology Development 40
Methodology • Cyc Method (Lenat & Guha, 1990) • Uschold & King (1995) • TOV Project (Gruninger & Fox, 1995) • Methontology (Fernandez-Lopez et al, 1997) • SWBP & Patterns (Rector, 2004) 41
Application-independent Modelling Generalisation or Super class MDA (Model Driven architecture) using UML-based modelling (Miller and Mukerji, 2003) 42
Protégé OWL Ontology Editor (Knublauch, 2003) 43
Using Reasoners in Classification Before classification: a Tree After: a Directed Acyclic Graph (Rector, 2004) 44
Jena-based Ontology Query Interface (George, 2006) 45
References BERNERS-LEE, T. , HENDLER, J. & LASSILA, O. (2001) The Semantic Web. Scientific American, 284(5), pp. 34 -43. BORST, W. N. (1997) Construction of Engineering Ontologies for Knowledge Sharing and Reuse. Ph. D. Thesis, SIKS - Dutch Graduate School for Information and Knowledge Systems. CHANDRASEKARAN, B. , JOSEPHSON, J. R. & BENJAMINS, V. R. (1999) What Are Ontologies, and Why Do We Need Them? IEEE Intelligent Systems, 14(1), pp. 20 -26. GEORGE, D. (2006) Developing Ontologies based on RDF-OWL Semantic Web languages [online]. Available from: dgeorge@uclan. ac. uk. [Accessed 13 April 2006]. GRUBER, T. R. (1993) A Translation Approach to Portable Ontology Specifications. Knowledge Acquisition, 5(2), pp. 199 -220. GUARINO, N. (1998) Formal Ontology and Information Systems. In: Proceedings of 1 st International Conference on Formal Ontologies in Information Systems (FOIS'98). Trento, Italy, 6 -8 June 1998. IOS Press, pp. 3 -15. KNUBLAUCH, H. (2003) An AI tool for the real world - Knowledge modeling with Protégé [online]. Java. World. Available from: http: //www. javaworld. com/javaworld/jw-06 -2003/jw-0620 -protege_p. html. [Accessed 23 December 2004]. LASSILA, O. & MCGUINNESS, D. (2001) The Role of Frame-Based Representation on the Semantic Web [online]. Technical Report KSL-01 -02, Knowledge Systems Laboratory, Stanford University, CA. Available from: http: //www. ep. liu. se/ea/cis/2001/005/cis 01005. pdf. [Accessed 12 July 2005]. LENAT, D. B. (1995) CYC: A Large-Scale Investment in Knowledge Infrastructure. Communications of the ACM, 38(11), pp. 32 -38. MILLER, J. & MUKERJI, J. (2003) Model Driven Architecture [online]. Object Management Group, Inc. Available from: http: //www. omg. org/docs/omg/03 -06 -01. pdf. [Accessed 29 September 2005]. RECTOR, A. , NOY, N. , KNUBLAUCH, H. , SCHREIBER, G. & MUSEN, M. (2004) Ontology Design Patterns and Problems: Practical Ontology Engineering using Protege-OWL [online]. Available from: http: //www. cs. man. ac. uk/~rector/tutorials/iswctutorial-2004/ISWC-Tutorial-Best-Practice. pdf. [Accessed 2 November 2005]. STUDER, R. , BENJAMINS, V. R. & D. FENSEL (1998) Knowledge Engineering: Principles and Methods. Data & Knowledge Engineering, 25(1 -2), pp. 161 -197. USCHOLD, M. F. & JASPER, R. J. (1999) A Framework for Understanding and Classifying Ontology Applications. In: Proceedings of the IJCAI-99 workshop on Ontologies and Problem-Solving Methods (KRR 5). Stockholm, Sweden, August 2 1999. pp. 11. 1 -11. 12. 46
Ontology Spectrum No specific hierachy Glossaries & Data Dictionaries Thesauri & Taxonomies Formal hierarchy & increasing expressiveness Formal Ontologies Inferencing (Lassila & Mc. Guinness, 2001, Uschold & Gruninger, 2004) 48
- Slides: 47