Meronymy and Inference
Traditional lexical semantic notions, like synonymy, hyponymy, antinomy, meronymy, etc., have now been explicitly encoded in databases, computational lexicons and semantic nets like WordNet. Large efforts have gone into constructing these resources and also into experimenting with methods for constructing them (semi-) automatically. But the questions related to how these resources can be exploited for inference has been less studied.
For hyponymy, the relationship to inference is simple: If A is a hyponym of B then every A is a B. But other notions, like meronymy (“part-of”) is not that simple. For example “finger” is a meronym of “hand”. This does of course not mean that “every finger is part of every hand”, but maybe something like “every finger is part of some hand”.
This project will consider the relationship between the meronymy-relation (between concepts) and the “part-of”-relation (between objects). What is the relationship between the two relations? Which properties do they have? How can this knowledge be used for textual inferences or other tasks in language technology? What kind of logic systems are appropriate (e.g. Horn clause logic, Description logic)?
The exact content of the project and which of the questions above it will address will depend on the background of the Student. It will be useful with some knowledge in semantics or logic beyond the obligatory courses.