Computational Models of Language Meaning in Context

This report documents the program and the outcomes of Dagstuhl Seminar 13462 “Computational Models of Language Meaning in Context”. The seminar addresses one of the most significant issues to arise in contemporary formal and computational models of language and inference: that of the role and expressiveness of distributional models of semantics and statistically derived models of language and linguistic behavior. The availability of very large corpora has brought about a near revolution in computational linguistics and language modeling, including machine translation, information extraction, and question-answering. Several new models of language meaning are emerging that provide potential formal interpretations of linguistic patterns emerging from these distributional datasets. But whether such systems can provide avenues for formal and robust inference and reasoning is very much still uncertain. This seminar examines the relationship between classical models of language meaning and distributional models, and the role of corpora, annotations, and the distributional models derived over these data. To our knowledge, there have been no recent Dagstuhl Seminars on this or related topics. Seminar 10.–15. November, 2013 – www.dagstuhl.de/13462 1998 ACM Subject Classification 1.2.7 Natural Language Processing