Towards ontology language handling imperfection

1
Towards ontology language handling imperfection

• Alan Eckhardt, Peter Vojtáš
• WIKT 2006

2
DL and Web modelling (cast prezentácie z ISWC06)

• DL basis for OWL
• Value restriction has to be reconsidered
• EL DL sufficient for applications
• EL querying in poly time
• Extension to user preference models
  e.g. a user looking for a cheap hotel close to a
  beach
• Aggregation of particular attribute preferences
  to global preference
• Top-k answers

3
Uncertainty in Web Modelling

• Probabilistic models in SW
• Fuzzy logic models SW
• Fuzzy EL - fuzzy concepts
• - crisp roles
• - fuzzy aggregation
• Enables concepts like
• _at_( price.cheap, distance.close)
• Problem of learning _at_ for each user

4
Description logic classical
AL
interpretation
EL
5
Description logic full fuzzification U.
Straccia
syntax
Fuzzy interpretation
Which t-norm t-conorm?
6
Bayesian networks (slide z clánku s M. Vomlelovou)

• Graph represents relations between variables
• Conditional probabilities

• Typical use
• We know A2, C2
• We ask for P(BA2,C2)

7
From FILP, IGAP to BN (slide z clánku s M.
Vomlelovou a T. Horváthom)
8
Bayesian EL DL and others

• Concepts are r. v. over preference scale
• Roles are crisp / certain
• Aggregation combination function
• Integrated with fEL, classical EL ,

9
Conclusion

• Sharing ideas
• Ideas supported by an analogy working in LP and
  simple experiments
• Further development of
• formal model
• experiments
• OWL extension
• Questions, comments,