Ontology and Search on the Semantic WEB

1
Ontology and Searchon the Semantic WEB

• SPINN PhD course
• PISA, September 2004
• Patrizia Paggio

2
Outline

• What is wrong with current search engines?
• The Semantic WEB Vision
• Semantic tagging
• Ontologies
• Practical assignment

3
What is wrong with current search engines?

• Answer to a search is a list of texts, but
• Is it the right texts?
• How about cross-lingual search?
• Do we always need texts?

4
Is it the right texts?

• Query Rom i dansk maleri fra 1800-tallet
• Rome in Danish painting from 1800
• Answer
• Dyreetiske Råd... Knudsen, Johannes Døden i Rom.
  ... Dansk kunst fra 1800-tallet i ... kunst,
  islamisk kunst fra ... Symbolismen i dansk og
  europæisk maleri ...
• (Committee for animal ethics... Knudsen, Johannes
  Death in Rome... Danish art i the 19th century
  in... art, Islamic art from... symbolism in
  Danish and European painting...)

5
Is it the right texts? (2)

• Syntax ignored
• no relation between Rom and dansk maleri in the
  text.
• Semantics ignored
• from maleri one should be able to get to malere
  (painters), billeder (paintings) etc.

6
Cross-lingual search (1)

• Query courses on 1800 century painting
• Answers
• University of Wisconsin-Madison
• Center for European Studies (Univ. Wisconsin)
• Rutgers, State University of New Jersey
• University of Dublin
• Enchanted Learning
• Only English-speaking, mostly American.

7
Cross-lingual search (2)

• Query courses on 1800 century painting at
  European universities
• Answers
• University of Wisconsin-Madison
• Elon University, North Carolina
• University College of London
• Institute of Art History in Marburg
• Cornell
• Still mostly English speaking, especially
  American!

8
Do we always need texts?

• Hvad er email-adresserne på alle forskere på KUA?
  
• (All emails of researchers from Humanities)
• We would like to see

What can we do now? web-page-ku.htm
9
Summary search engines

• More syntax and semantics
• Extraction of knowledge from texts and databases
  rather than just text retrieval
• Extraction of knowledge from texts on different
  languages

10
The Semantic WEB

• A more intelligent WEB in which information
  (documents and other data) are tagged so that
  computer application understand them, or know
  how to use them.
• 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.
• (Berners-Lee, Hendler and Lassila 2001)

11
Semantic tagging (1)

• HTML documents are tagged so that browsers
  understand their syntax and show them to users
  who interpret them.
• lth3gt Patrizia Paggio lt/h3gt
• ltpgt Employee at lta href"http//cst.dk"gt CST lt/agt
  in Copenhagen lt/pgt
• ltpgt Titel Seniorforsker lt/pgt
• ltpgt Email patrizia_at_cst.dk lt/pgt

12
Semantic tagging (2)

• Semantic metadata make the content of WEB pages
  explicit.
• hasName (SeniorResearcher, Patrizia Paggio)
• hasName (ResearchCentre, CST)
• Affiliated (Patrizia Paggio, CST)
• subClass (SeniorResearcher, employee)
• hasProperty (Patrizia Paggio, emailpatrizia_at_cst.d
  k)
• hasProperty (CST, urlhttp//cst.dk)

13
Defining semantic tagsConcept hierarchies
Employee
Faculty
Administrative
Researcher
Assistent
Professor
Lecturer
Senior Researcher
14
Adding relations
15
Ontologies What are they?

• A domain ontology defines concepts and relations
  between them for a specific domain.
• It provides the semantic vocabulary for semantic
  tagging on the semantic WEB.
• For domain ontologies to be useful, they must
  conform to standards (W3C).
• Examples XML, RDF, OIL, DAML-OIL, OWL, Topic
  Maps.

16
Ontologies - Repositories

• DAML Ontology library
• 282 ontologies
• total no. of classes 67987
• total no. properties 11149
• total no. of instances 43646

17
Ontology-based querying

• Hvad er email-adresserne på alle forskere på KUA?
• will retrieve a list of emails and researchers
  provided that
• relevant pages are tagged semantically
• current standards are used
• the agent (search engine) knows the underlying
  ontology or can find it.

18
Ontologies Some issues

• Some crucial issues
• How are ontologies produced?
• How is the semantic tagging produced?
• Will knowledge referring to the same domain
  always be described by the same ontology? What if
  it doesnt?
• How is knowledge expressed in different languages
  related to the same domain ontology?

19
Practical assignment (1)

• Define ontologies for web pages in different
  languages. Find
• Classes Organisation
• Subclasses Institute is-a Organisation
• Instances Person Patrizia Paggio
• Attributes Email (Person)
• Relations Teach (Teacher, Course)

20
Practical assignment (2)

• (2) Define correspondences between the
  ontologies.
• Ontology I Relation Ontology II
• Person Equivalent Person
• Professor Similar Professor
• (3) Discuss the principles followed to establish
  the correspondences.