Lunch seminar: Building a Computational Grammar of English

This autumn's lunch seminar from the CAS-project SynSem: From Form to Meaning - Integrating Linguistics and Computing is held by Senior Researcher Dan Flickinger (Center for the Study of Language and Information, Stanford University).

Enabling computers to process natural language in meaningful ways presupposes formalized knowledge about grammar, in particular about the lexicon and system of rules that human language users put to use effortlessly, to spontaneously generate and comprehend previously unheard of utterances from an in principle infinite set of possible linguistic structures.

The SynSem group combines researchers from linguistics and informatics, and we will introduce key themes from our work through a representative (albeit small) sample.

Dan Flickinger has worked on implementing a comprehensive grammar of English since the early 1990s, using a specialized linguistic knowledge representation language and software that can parse and generate, i.e. compute the correspondences between surface strings (linguistic signals) and associated logical forms (meanings). The grammar today can successfully analyze around 90 % of all utterances in a broad variety of running texts, and it is being applied commercially (among other things) in second-language tutoring.

At SynSem, Dan pursues theoretical and practical questions related to this work. For example, how to assign a meaning representation to so-called gapping constructions: ‘Oda ordered sushi for Joakim and curry for Mary.’  Or the question of exactly how many people were involved in the situation described by ‘five men and women got married today’?

 

For background, please see: http://erg.delph-in.net

 

Please e-mail CAS Oslo in advance if you want to attend: cas-info@cas.oslo.no