Presentation is loading. Please wait.

Presentation is loading. Please wait.

11/8/20051 Ontology Translation on the Semantic Web D. Dou, D. McDermott, P. Qi Computer Science, Yale University Presented by Z. Chen CIS 607 SII, Week.

Similar presentations


Presentation on theme: "11/8/20051 Ontology Translation on the Semantic Web D. Dou, D. McDermott, P. Qi Computer Science, Yale University Presented by Z. Chen CIS 607 SII, Week."— Presentation transcript:

1 11/8/20051 Ontology Translation on the Semantic Web D. Dou, D. McDermott, P. Qi Computer Science, Yale University Presented by Z. Chen CIS 607 SII, Week 7

2 11/8/20052 Overview  Main Content: Ontology Translation  Ontology Merging and Automated Reasoning  Dataset translation, extension generation and querying  Motivation  Web agent should understand and process web data  Ontology: formalization of web contents (voc, axiom)  Ontologies are very different  Proposed Solution: Ontology Translation  Step 1: Ontology merging: union of terms and axioms  Step 2: Bridging axioms are manually added  Step 3: Automatic reasoning with theorem prover

3 11/8/20053 Why Ontologies Are Different?  Syntactically Different  DAML+OIL, OWL, WSDL, etc.  Semantically Different  Different taxonomic structures of concepts. Examples: firstname/lastname and full name Yale’s term for Article, Inproceedings, Incollection, and CMU’s term for Article Semantic difference can be inherited (birth inherited event)

4 11/8/20054 Ontology Translation Problems  Dataset Exchange Problem  To exchange information between ontologies  Ontology Extension Generation  Given O1, O2 and O1’s extension O1s, construct O2s.  Example: DAML-S (app), WSDL (protocol), Congo and/or BravoAir (extend DAML-S), construct ontology in protocol level (extend WSDL)?  Query from Multiple Ontologies  Knowledge may be in multiple knowledge bases

5 11/8/20055 Closely Related Work  Difference with Ontology Mapping  Automatic discovery of mapping rules (correspondence)  Unlikely to be fully automatic due to: Accuracy, Complexity of mapping rules  Ontology translation is based on a small set of axioms (question: bridging axiom vs. mapping rules)  Ontolingua  Any ontology from/to a “generic” ontology  Unlikely to scale well  OntoMorph  Case-by-case translation (dataset-by-dataset)  Not general methodology; always case-by-case  A Small Summary: Automation in the *right* level

6 11/8/20056 Dataset Translation

7 11/8/20057 Three Examples of Web-PDDL

8 11/8/20058 Semantic Translation  Problem: Given a set of facts in one vocabulary, infer the largest possible set of consequences in another.  Merge:  Union of terms and axioms (automatic? The paper says manual construction by experts).  Adding bridging axioms (at best semi-automatic) Relate symbols in one ontology to symbols in another.

9 11/8/20059 Merge: Bridging Axiom Figure 2. A Bridging Axiom Figure 3. Term generating functions

10 11/8/200510 Inferences  Inference Engine Is Used for:  Forward chaining to reform facts  Backward chaining to reform query  Introducing term-generation functions

11 11/8/200511 Ontology Extension Generation  Problem: O1, O2 and O1s, how about O2s?  Similar to Dataset Translation  Take the two ontologies as source and target  Take the extended ontologies as fact dataset  Use inference engine to generate translated facts  Create new predicates for the translated facts and make them subproperties of the predicates in the conclusion.  Then generate the corresponding axioms for subproperty relationships  Evaluations  Only translate predicates, types and axioms about subproperties; not currently working for more general axioms.

12 11/8/200512 Query through Different Ontologies  Problem: Knowledge from different ontologies, or even unable to translate the query by a single ontology  Steps: query selection and query reformulation  Query selection: chose a simple one and got answered by an (potentially merged) ontology  Query reformulation: backward chaining to reform the rest subqueries and get another seleciton  Evaluations  Query optimization is not done yet

13 11/8/200513 Questions  Why specifying bridging axioms is easy?  Evaluation of inference engine? Completeness problem? What is the best logic?

14 11/8/200514 Thank you! Presented by Zebin Chen CIS 607 SII, Week 7


Download ppt "11/8/20051 Ontology Translation on the Semantic Web D. Dou, D. McDermott, P. Qi Computer Science, Yale University Presented by Z. Chen CIS 607 SII, Week."

Similar presentations


Ads by Google