Robust Computational Semantics

Practical tasks like question answering and machine translational ultimately require computing meaning representations that support inference. Standard linguistic accounts of meaning are impracticable for such purposes, both because they assume nonmonotonic operations such as quantifier movement, and because they lack a representation for the meaning of content words that supports efficient computation of entailment. I’ll discuss practical solutions to some of these problems within a near-context free grammar formalism for a working wide-coverage parser, in current work with Mike Lewis, and show how these solutions can be usefully applied in NLP tasks. Mark Steedman. 2013. Robust Computational Semantics. In Proceedings of Australasian Language Technology Association Workshop, page 2.