# ISWC2005 Notes: Uncertainty Reasoning for the Semantic Web 3

During the coffee break I managed to locate a power socket, got my mouse out and found some sustenance - I'm ready for the lightning talks now!

- Modeling Degrees of Conceptual Overlap in Semantic Web Ontologies - ontologies are based on crisp logic and do not procvide means foor expressing degrees of subsumption. Partial overlap of concepts, e.g. Lapland spans several countries. Goal is to determine relevance of hits based on conceptual overlap.
- Modeling the Non-Expected Choice: A Weighted Utility Logit - we commonly expect to apply the von Neumann-Morganstein expected utility theory in our choice models. We frequently violate the same theory in our everyday choices. Their remedy (sorry I missed it!) allows better predictions about how real users are going to evaluate different probability distributions of choice alternatives. Allows the modeler/user of a search agent to specify personalities, possiuble helpful in classification tasks.
- Ontology Based Analysis of Experimental Data - interesting computer metaphor of DNA (what applications can run), mRNA (what is running in our body), Proteins (what is our body actually doing). Language to define scoring functions on ontologies and data. Evaluate consistency between ontologies and data.
- Paraconsistent Reasoning for the Semantic Web - classical logic requires complete absence of inconsistencies. Paraconsistent logics challenge "ex contradictione quodlibet" rule ("from a contradiction every proposition may be deduced"). Allow sensible reasoning even in presence of inconsistencies. Use cases: distributed information systems, coping with change, different opinions, dialetheias (liar's paradox). Approaches - relevant logics (Anderson and Beinap), many-valued systems (logic with more than two truth values), non-adjunctuve systems (do not allow to conclude A from A ∧ B)Jaskowski), non-truth-functional logics (da Costa)
- Representing Probabilistic Relations in RDF - why use probability for uncertainty? Probability is intuitive, well studied mathematically. What kind of prob. relations are to be represented? Both instance (A-box) and class level (T-Box). Why and how should they be represented in RDF? to keep RDF as basic representation language in semweb.Not by using quads but using a core vocab. Probabilities are attached to a partition. A prob:Clause is similar to reification. Can be cumbersome, perhaps better to introduce higher language.

Wow, some of those were really quick.