Proceedings of ALTSS/ALTW, Melbourne, December 2003
Deep Lexical Acquisition
Timothy Baldwin, CSLI Stanford
ABSTRACT:
Deep processing involves applying "precision grammars"
(i.e. linguistically-precise grammars, such as HPSGs) to natural language
analysis, and has significant advantages over shallow methods in terms of its
ability to capture fine-grained lexical and constructional interactions and
produce a rich semantic representation. The main limitation of deep processing
is coverage, which tends to be restricted due to the detailed annotation
required to encode individual lexical items in precision grammars. This talk
will tackle the question of how to expand the coverage of a precision grammar
through the automatic acquisition of lexical features and ultimate type
classification of a given word. I will use the English Resource Grammar as a
test case, and outline a range of methods by which new lexical items can be
acquired either directly through the application of the grammar, or indirectly
through techniques drawing on corpus data and/or semantic ontologies.
BIO :
Timothy Baldwin is a Senior Researcher at the Centre for the Study of Language
and Information (CSLI), Stanford University. He is a member of the CSLI LinGO
Multiword Expression Project, specialising in the lexical acquisition,
semantic classification and machine translation of multiword
expressions. Other recent research interests include computational lexical
semantics, the interface between theoretical and computational linguistics,
and computer-assisted language learning applications for computational
linguistics.[http://www-csli.stanford.edu/~tbaldwin/]
RESOURCES:
Materials for this lecture are available from http://www.alta.asn.au/events/