1 / 20

Enriching OWL with Instance Recognition Semantics for Automated Semantic Annotation

Enriching OWL with Instance Recognition Semantics for Automated Semantic Annotation. Stephen W. Liddle Information Systems Department Yihong Ding & David W. Embley Computer Science Department Brigham Young University, Provo, Utah. Outline. Background Instance recognition semantics

teenie
Télécharger la présentation

Enriching OWL with Instance Recognition Semantics for Automated Semantic Annotation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Enriching OWL with Instance Recognition Semantics for Automated Semantic Annotation Stephen W. Liddle Information Systems Department Yihong Ding & David W. Embley Computer Science Department Brigham Young University, Provo, Utah

  2. Outline • Background • Instance recognition semantics • Automated semantic annotation • OWL-AA: OWL for Automated Annotation • Related work • Conclusion ONISW 2007, Auckland, New Zealand

  3. Motivation • Big Hairy Audacious Goal: deliver a machine-processable semantic web that will let us build agents to work intelligently on our behalf • Realistic Shorter-term Goal: provide automatic semantic annotation machinery that turns ordinary web pages into semantic web pages by adding appropriate metadata • Goal of this Paper: Show that we can extend OWL with epistemological declarations that let us use our data extraction engine to provide a solid automatic annotator that works with standard ontologies ONISW 2007, Auckland, New Zealand

  4. Typical Approach • Typical approach to automatic annotation: • Use existing data extraction engine to wrap and annotate pages • BUT these approaches do not match extracted data with an ontology • Main drawback: • Post-processing step to map extracted data to the ontology ONISW 2007, Auckland, New Zealand

  5. Our Approach • Our data extraction engine is ontology based and does not need a post-processing alignment step • We have demonstrated that our approach works (Ding/Embley/Liddle 2006, LNCS 4185) • But our ontology language is not a W3C standard • OWL is a standard (W3C recommendation) • But lacks sufficient declarative instance-recognition semantics for automatic annotation ONISW 2007, Auckland, New Zealand

  6. A Solution • Extend OWL with declarative instance-recognition semantics • Call this OWL for Automated Annotation (OWL-AA) • OWL-AA extension is epistemological in nature • Ontological definitions should be independent of the form of knowledge representation • Epistemological definitions may be sensitive to the form of knowledge representation ONISW 2007, Auckland, New Zealand

  7. Our Contributions • OWL-AA extends OWL to provide for automated semantic annotation that • Embeds instance-recognition semantics declarations in ontologies and data extraction tools • Provides enhanced knowledge sharing and reuse via semantic web • Separates domain knowledge creation from semantic annotation process • Thus maintaining separation between concerns of domain experts and software developers ONISW 2007, Auckland, New Zealand

  8. Instance Recognition Semantics • Instance Semantics Recognizers (ISR’s) • Formal specifications that identify instances of a concept in unstructured, semistructured, or structured text • Example: • Concept: PhoneNumber • Text: “Call me at 555-1212.” • ISR should recognize that 555-1212 in the text has the intensional meaning of PhoneNumber. ONISW 2007, Auckland, New Zealand

  9. ISR Declarations in Data Frames BedroomNr internal representation: Integer external representation: [1-9]|10 left context phrase: \b right context phrase: .*r(oo)?ms? exception phrase: \s.*ba(th)?s?\b.*r(oo)?ms? context keywords: b(r|d)s?|bdrms?|bed(rooms?)? ... LessThan(p1: Price, p2: Price) returns (Boolean) context keywords: less than|<|or less|fewer| ... ... end Feature external representation: ApartmentFeature.lexicon ... end ONISW 2007, Auckland, New Zealand

  10. Resiliency • There are no document layout dependencies in our ISR declarations • Other approaches typically use page layout information for extraction • Layout independence (resiliency) is a major benefit • Page layouts change periodically on the web • Different publishers in the same domain generally use different page layouts • Rewriting wrappers is costly ONISW 2007, Auckland, New Zealand

  11. External Representation Context Phrase External Representation Context Phrase External Representation External Representation Context Phrase Context Keyword External Representation Annotation (Apartment Rental) BedroomNr BathNr CAPITOL HILL Luxury 2 bdrm 2 bath, 2 grg, w/d, views, 1700 sq ft. $1250 mo. Call 533-0293 Feature MonthRate ContactPhone ONISW 2007, Auckland, New Zealand

  12. ISR-Based Annotation Demo http://www.deg.byu.edu, demos section ONISW 2007, Auckland, New Zealand

  13. hasException P ISRvalue P extractionLeftContext P extractionPattern P extractionKeyword extractionRightContext P P RDFS Graph of OWL-AA domain domain ISR subClassOf subClassOf range subClassOf range &xsd;string ExternalRepresentation ContextualRepresentation Exception subClassOf subClassOf subClassOf subClassOf RegularExpression ContextPhrase ContextKeyword LexiconList range subClassOf subClassOf LeftContextPhrase RightContextPhrase range range domain range domain domain domain owl:Class ONISW 2007, Auckland, New Zealand

  14. OWL-AA Details (1) • Add namespace xmlns:owlaa=“http://www.deg.byu.edu/OWL-AA#” • Add properties to OWL classes associating ISR declarations <owl:onProperty rdf:resource=“owlaa:extractionPattern” /> <owl:hasValue rdf:resource=“BedroomNr-1” /> ONISW 2007, Auckland, New Zealand

  15. OWL-AA Details (2) • Declare ISR value and contextual restriction <owlaa:RegularExpression rdf:ID=“BedroomNr-1”/> <owlaa:ISRvalue rdf:datatype=“&xsd;#string”> [1-9]|10</owlaa:ISRvalue> <owlaa:extractionLeftContext rdf:resource=“#leftContext-1”/> <owlaa:extractionRightContext rdf:resource=“#rightContext-1”/> <owlaa:extractionKeyword rdf:resource=“#contextKeyword-1”/> </owlaa:RegularExpression> ONISW 2007, Auckland, New Zealand

  16. OWL-AA Details (3) • Add context phrases, context keywords, and exceptions <owlaa:LeftContextPhrase rdf:ID=“leftContext-1”/> <owlaa:ISRvalue rdf:datatype=“&xsd;#string”> \b</owlaa:ISRvalue> </owlaa:LeftContextPhrase> ... (see our paper for the rest of the details) ONISW 2007, Auckland, New Zealand

  17. OWL-AA Features • Syntactically, OWL-AA is attachment-independent w.r.t. OWL • No need to change a single line of existing OWL ontologies • Semantically, OWL-AA is also attachment- independent w.r.t. OWL • ISR declarations do not change meaning of existing ontology concepts at conceptual level • Prototype implementation translates OWL-AA to our own internal format for processing ONISW 2007, Auckland, New Zealand

  18. Related Work • We aren’t the first to work on ISR declarations • IBM’s Unstructured Information Management Architecture • There are many proposals for extending OWL • Two extensions related to ours • Context OWL (C-OWL) • Extension for ontology mapping that localizes ontology content for global visibility • OWL-Eu • Extension that enriches OWL with customized datatypes, a significant need • Not attachment-independent, unfortunately • But our approach is compatible with OWL-Eu too ONISW 2007, Auckland, New Zealand

  19. Conclusion • Automated semantic annotation is an important and fundamental problem for the semantic web • Key to automated annotation is ISR declaration • Our epistemological OWL-AA extension augments OWL to formalize ISR declarations ONISW 2007, Auckland, New Zealand

  20. Conclusion (cont.) • OWL-AA is fully compatible with ordinary OWL • Attachable/detachable • OWL-AA does not introduce complexity or decidability issues into OWL • Prototype implementation demonstrates that our approach works well • Embedding epistemological declarations in OWL is a fruitful area meriting further research ONISW 2007, Auckland, New Zealand

More Related