Transcript View/Download

OWL-AA: Enriching OWL with
Instance Recognition Semantics for
Automated Semantic Annotation
2006 Spring Research Conference
Yihong Ding
Semantic Web and
Automated Semantic Annotation
 Semantic Web: the web containing machine-
processable web data
 Semantic Annotation: adds formal metadata to web
pages


Metadata links data in a web page to defined concepts
in an ontology
Annotated data becomes machine-processable
 Annotation needs automation to be scalable
2
“Main Drawback” of Current
Automated Semantic Annotation
 Problem: “post-processing and mapping of the IE
[information extraction] results to an ontology”
[Kiryakov 2004]


Needs human intervention
Decreases system automation and scalability
 Solution: “use ontolog[ies] more directly during the
process of extraction” [Kiryakov 2004]


Does work (as our ontology-based annotation shows)
But …
3
A Hidden Problem:
Compatible with Standards
 A solution should be compatible with semantic web
standards


OWL (Web Ontology Language): standard
Solutions must be OWL-compatible
 Current Solution
 OSMX (Object-oriented Systems Model in XML): not a
standard, not OWL-compatible
 Declarative instance recognition semantics
 Needed by automated annotation process
 Lacking in OWL
4
Instance Recognition Semantics in
Extraction Ontologies
 Instance recognition semantics: machine-
processable recognizers of instances that belong to
the extention of a concept in a specified domain.
 Examples in extraction ontologies
 External Representation
 Price: \d+|\d?\d?\d,\d\d\d
 Make: CarMake.lexicon
 Contextual Representation
 Context phrases (left, right), e.g. \$?
 Context keywords: e.g. price | obo | neg(\.|otiable)
5
OWL: Lacks Instance Recognition
Semantics
 In general, OWL



Declares class, property, hierarchical relationship, restriction.
Declares instantiations.
Does not support declaration of “instance recognition”
 Consequently,


Not enough declarative semantics in OWL directly useable
by automated annotation
Mixture of knowledge declaration and knowledge processing



Domain experts must know program implementation;
Or, program developers must be domain experts.
No annotation integrity checking


<carad:Make>Taurus</carad:Make> is legal, though it is
incorrect;
And, machines cannot catch this error.
6
OWL-AA (RDF Schema)
7
OWL-AA (RDF Schema)
8
OWL-AA Declarations
9
Implementation
 Jena API converts OWL-AA ontologies to
OSMX ontologies
 Use OSMX ontologies to do automated
annotation
10
Conclusion
 OWL-AA is a way to extend OWL to provide for
automated semantic annotation.
 OWL-AA overcomes the “main drawback” of
automated semantic annotation.
 OWL-AA allows us to separate the creation of domain
knowledge from the implementation of a processor to
use domain knowledge for the purpose of annotating
web pages.
 OWL-AA provides for annotation integrity checking.
11
Declaration vs. Instantiation
Instantiation
Instantiation
Declaration
12
Instance Recognition Semantics
Machine-processable recognizers of
instances that belong to the extention of a
concept in a specified domain.
IRecS of Left
Concept:
has line in eye
IRecS of Right
Concept:
no line in eye
13