Download presentation
Presentation is loading. Please wait.
Published byBrittany Manning Modified over 8 years ago
1
Ontology Technology applied to Catalogues Paul Kopp
2
Paul KoppWGISS22Annapolis 2 About Descriptive Logics ■Descriptive Logics is a formalism for representing knowledge ■Knowledge concerns concepts, roles and individuals An individual is in relationship with at least some concept Example “Aristotle” is an individual “human being” is a concept “is-a” is a relation The individual “Aristotle” is in the relationship “is-a” with the concept “human being” A concept is actually a set of individuals Concepts may be built by applying operations to atomic concepts Examples: restriction, intersection, union A role is a relationship between two concepts Example “hasChild” is a role expressing the relationship between the concepts “parent” and “child” A role is actually a relationship between individuals Constructions and Restrictions may apply to roles as well (ex.: intersection, “≥3hasChild”)
3
Paul KoppWGISS22Annapolis 3 About Descriptive Logics (continued) ■Knowledge Representation Systems based on Descriptive Logics Descriptions TBox (Terminology Box) –Contains the terminology given as concepts, roles and constructs on them –Ex.: (a) “satellite” “is-a” “space engine” –A TBox can contain complex descriptions, depending on the constructors that are available (i.e. on the definition of the Description Language) ABox (Assertion Box) –Contains assertions about individuals in relationship with the terminology –Ex.: “SPOT5” “is-a” “Satellite” (SPOT5 is an individual) Reasoning Services From the TBox one infers properties of descriptions –Satisfiability, Subsumption, Equivalence, Disjointness From the ABox, one infers properties of assertions w.r.t. the TBox –Consistency, Retrieval (find all individuals that are instances of a given concept), Realization (find the most appropriate concept for a given individual) Complexity of reasoning is a major issue –Rich Description Languages entail complex reasoning (with possible untractability of some inferences like subsumption)
4
Paul KoppWGISS22Annapolis 4 About Descriptive Logics (continued) ■Comparison with other Knowledge Representation Formalisms Conceptual Graphs (Sowa) Semantic Data Models Entity Relationship Model (Chen) Conceptual Modelling Unified Modelling Language (Rumbaugh) Specialists have been studying the correspondences between all these formalisms.
5
Paul KoppWGISS22Annapolis 5 About ontologies ■From ontos (Greek οντοσ = “which is real”) and logos (Greek λογος = “word”, “speech”) ■Name given to knowledge representations where the main relationship is the “is-a” relationship ■Ontologies are used to describe the concepts that prevail in a domain ■Thesauri are very simple ontologies Excerpt from the IDN keywords: ATMOSPHERE >ATMOSPHERIC WATER VAPOR >EVAPOTRANSPIRATION ATMOSPHERE >ATMOSPHERIC WATER VAPOR >EVAPOTRANSPIRATION ■Tools to create ontologies Racer (http://www.racer-systems.com)http://www.racer-systems.com Stands for Renamed ABox and Concept Expression Reasoner Protege (http://protege.stanford.edu)http://protege.stanford.edu SWOOP (http://www.mindswap.org/2004/SWOOP)http://www.mindswap.org/2004/SWOOP ■Reasoners Pellet (http://www.mindswap.org/2003/pellet)http://www.mindswap.org/2003/pellet Racer
6
Paul KoppWGISS22Annapolis 6 Ontologies and the W3C ■Web Ontology Language (OWL) Defined by the W3C Specification of ontologies using the xml/RDF schema Concept = class in OWL Role = property in OWL 3 levels of expressiveness OWL-Lite (for simple ontologies like thesauri) OWL-DL (for ordinary Descriptive Logics compliant ontologies) OWL-Full (no computational guarantee) ■SPARQL Query Language for RDF Defined by the W3C Specification of queries on ontologies specified with OWL ■OWL and SPARQL implemented in several ontology tools (ex.: Protege)
7
Paul KoppWGISS22Annapolis 7 Ontologies and Catalogues ■Catalogue primary entries are metadata ■Traditional metadata retrieval Queries are applied to predefined “queryable” metadata elements (title, keyword, etc.) Metadata satisfying the query are retrieved from the database and presented to the user (full or predefined “brief” or “short” content) ■Another way to retrieve metadata? Express metadata as ontologies Link the metadata expressed as ontologies to a reasoner Make the reasoner available to the catalogue user The catalogue user may ask any question he wants to the catalogue
8
Paul KoppWGISS22Annapolis 8 Experiment at CNES ■Extension of the CNES ISO19115 Metadata Catalogue Developed under the auspices of CNES R&D (project R-S06/OT-0005-013) Metadata are inserted as ordinary xml files Metadata may also (additionally) be inserted as ontologies (OWL-DL) The catalogue user queries the catalogue as usual (keyword, time, location, etc.) The catalogue user may ask for “more queries” The catalogue system opens a reasoner (Pellet, through its Java API) The user prepares a query from predefined SPARQL templates (the user just enters the values for the query variables) SPARQL templates are prepared by the Catalogue Manager Results from the “more queries” function are merged with the previous ones and presented to the user ■End of development expected in 4Q06
9
Paul KoppWGISS22Annapolis 9 Thank you!
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.