Cog-D: Distributional Semantics* (Advanced)(Denis Paperno, CNRS/University of Lorraine)
This Seminar is advanced. Some background in formal semantics is required.
The course introduces the distributional approach to semantic modeling. It is increasingly popular in computational and cognitive research and has even been proposed as a solution to ‘Plato’s problem’ (Landauer and Dumais 1997). In the distributional approach, semantic representations of linguistic units are derived from the statistics of their usage in text.
Topics to be discussed in the seminar include :
• creation of distributional models and their parameters ;
• psycholinguistic applications of distributional models (modeling semantic similarity and relatedness judgements, priming effects, and cognitive biases) ;
• applications to syntactic preferences ;
• compositionality in distributional semantics ;
• applications to morphology ;
• extracting semantic properties from distributional representations ;
• application to pragmatics ;
• applications to language typology and to historical linguistics ;
• the problem of grounding : distributional semantics and computer vision.