Bell Labs logo

The Semantic Web and (vs?)
Knowledge Representation

Peter F. Patel-Schneider, pfps@research.bell-labs.com

Abstract

The Semantic Web has been attracting considerable attention the last few years. From the point of view of Knowledge Representation, the Semantic Web affords opportunities for both research and application. However, several aspects of the Semantic Web, as it has been envisioned, cause problems from the Knowledge Representation viewpoint. Overcoming some of these problems has resulted in a more formal basis for the Semantic Web and an increase in expressive power in Semantic Web languages. Other of these problems still remain and need a new vision of the Semantic Web from a Knowledge Representation viewpoint.

Recent Activities

This talk ties these activities into a (fairly provocative) whole and presents a vision of a potential future Semantic Web. I will not have time to detail everything mentioned in the talk and will be making some simplifications (that should not affect the truth of the message).

Underlying Theses

Thesis 1: Formal symbolic representation of information has use.

Evidence for: Databases
Evidence against: Information Retrieval vs Natural Language Understanding,
                                Modern Maching Learning methods

Thesis 2: The Semantic Web is a good place for formal symbolic representation.

The Semantic Web is an extension of the current web in which information is given well-defined meaning, better enabling computers and people to work in cooperation.
[Tim Berners-Lee, James Hendler, and Ora Lassila, "The Semantic Web", Scientific American, 284(5), May 2001, pp. 34–43]

Why the Semantic Web?

So, what does it take (technically) to make the Semantic Web a good place for formal symbolic representation?

What is the Semantic Web?

Semantic Web Stack

Initial Version of the Semantic Web (1990s)

Vision: The Semantic Web is an extension of the current web in which information is given well-defined meaning.

Providing Meaning for RDF and RDFS

The One-Language Vision of the Semantic Web

Vision: One language to rule over them.

  1. All information in the Semantic Web is written in RDF/XML.
  2. All information in the Semantic Web is carried as RDF triples.
  3. Triples always carry (at least) their RDF meaning.

Benefits of vision:

Status of vision:

Adding a Web Ontology Language

OWL Web Ontology Language

Wine ≤ PotableLiquid ∩ (=1 hasMaker) ∩ (∀ hasMaker Winery)
                     ∩ (≥ 1 madeFromGrape) ∩ (= 1 hasColor)
                     ∩ ...
madeFromGrape ≤ Wine × WineGrape
hasColor ≤ Wine × WineColor
WineColor = { White, Rose, Red }
White ≠ Rose   White ≠ Red   Rose ≠ Red
WineDescriptor = WineTaste ∪ WineColor
 
WhiteWine = Wine ∩ ( hasColor : White )
Riesling = Wine ∩ (madeFromGrape : RieslingGrape )
                ∩ (≤ 1 madeFromGrape)
Riesling ≤ hasColor : White
 
CorbansDryWhiteRiesling ∈ Riesling ∩ (hasMaker : Corbans) ∩ ...

Taken from the OWL wine and food ontologies. This example uses a compact syntax often used in papers about Description Logic.

Various syntaxes for OWL

Description Logic Publication Syntax:

WhiteWine = Wine ∩ ( hasColor : White )

Functional-style (abstract) Syntax:

Class(x:WhiteWine complete x:Wine
       restriction(x:hasColor value(x:White)))

RDF/XML:

<owl:Class rdf:ID="WhiteWine">
  <owl:intersectionOf rdf:parseType="Collection">
    <owl:Class rdf:about="#Wine" />
    <owl:Restriction>
      <owl:onProperty rdf:resource="#hasColor" />
      <owl:hasValue rdf:resource="#White" />
    </owl:Restriction>
  </owl:intersectionOf>
</owl:Class>
      

RDF Triples:

x:WhiteWine rdf:type owl:Class .
x:WhiteWine owl:intersectionOf _:l1 .
_:l1 rdf:type rdf:List .
_:l1 rdf:first x:Wine .
_:l1 rdf:next _:l2 .
_:l2 rdf:first _:r1 .
_:l2 rdf:next rdf:nil .
x:Wine rdf:type owl:Class .
_:r1 rdf:type owl:Restriction .
_:r1 owl:onProperty x:hasColor .
_:r1 owl:hasValue x:White .

Characteristics of OWL

Fitting OWL into the Semantic Web

Three problems:
Triples, triples, and TRIPLES.
  1. Triples are the only syntax.
  2. Triples are (always) facts.
  3. Triples don't distinguish between uses of names:
    • No distinction between individual, class, and property names.
    • No distinction between domain properties and built-in properties.

Solutions to these problems consumed much of the efforts of the W3C Web Ontology working group.
[Peter F. Patel-Schneider and Dieter Fensel. "Layering the Semantic Web: Problems and Directions". First International Semantic Web Conference. Sardinia, Italy, June 2002.]
[Ian Horrocks, Peter F. Patel-Schneider, and Frank van Harmelen. "From SHIQ and RDF to OWL: The Making of a Web Ontology Language". Journal of Web Semantics 1(1), December 2003, pages 7–37.]

OWL Syntax as Triples

OWL Classes as Objects

OWL Syntax Connectives as Properties

OWL DL: A Safe and Sane Version of OWL

(Aside) What you can't do in OWL DL

  1. Fiddle with the OWL syntax connectives.
    • owl:intersectionOf rdfs:subPropertyOf owl:unionOf .
    • Modifies meaning of OWL syntax.
    • No one in their right mind should even think of doing this!
  2. Create bits of OWL syntax with missing or extra parts.
    • _:x rdf:type owl:Restriction .
      _:x owl:onProperty ex:loves .
      _:x owl:minCardinality "1"^^xsd:nonNegativeInteger .
      _:x owl:allValuesFrom foaf:Person .
    • Such syntax has unusual and wide-ranging consequences.
    • No one in their right mind should even think of doing this!
  3. Use a name as both a class and an individual, or both a class and a property, or ....
    • _:y rdf:type ex:Eagle .
      ex:Eagle rdf:type us:EndangeredSpecies .
    • OWL DL thus does not support meta-modelling.
    • Something that is sometimes wanted.
      • There are ways to get something similar.
      • OWL 1.1 (a new version of OWL, in development) will have punning, which supports meta-modelling.

Fitting First-Order Logic into the Semantic Web

What makes the Semantic Web
the Semantic Web?

The vision is broken, so we need to change it somehow.

What should remain?

My Vision of the Semantic Web

What Comes Next

What I think is needed to make the Semantic Web more useful, and what I would like to do (in no particular order).

Lots of work here, some medium-term and some long-term.

Other References

Peter F. Patel-Schneider and Jerome Simeon. Building the Semantic Web on XML. Best paper award winner, First International Semantic Web Conference. Sardinia, Italy, June 2002.

Peter Patel-Schneider and Jerome Simeon. The Yin/Yang Web: A Unified Model for XML Syntax and RDF Semantics. IEEE Transactions on Knowledge and Data Engineering: Special Issue on WWW2002 15, 3, July/August 2003, pages 797–812.

Ian Horrocks and Peter F. Patel-Schneider. Three Theses of Knowledge Representation in the Semantic Web. The Twelfth International World Wide Web Conference. Budapest, Hungary, May 2003, ACM Press, pages 39–47.

Bijan Parsia and Peter F. Patel-Schneider. Meaning and the Semantic Web. The Thirteenth International World Wide Web Conference. New York, New York, May 2004, ACM Press.

Peter F. Patel-Schneider. A Revised Architecture for Semantic Web Reasoning. Third Workshop on Principles and Practices of Semantic Web Reasoning. Dagstuhl, Germany, September 2005. LNCS 3703, Springer Verlag, 2005.