ESSLLI'2000 Workshop on
Linguistic Theory and Grammar Implementation
Invited talk on 14. August 2000
Martin Kay:
Taking Underspecification Seriously
Abstract
A major attraction of unification-based formalisms is that they offer
the possibility of treating vagueness differently from ambiguity. We
do not have to say that "sheep" has two lexical entries, one for the
singular and one for the plural; we can give it one entry that leaves
its number unspecified. But as more categories are involved in the
syncretism, unification no longer provides an easy solution. So
German adjectives in -en, French first-conjugation verbs in -e, and
innumerable other cases of morphological neutralization, give rise to
large numbers of variants for a word which parsers then multiply by
large numbers of variants for other words to give an exponential
result. One answer to the problem that this poses is to bring in
heavy machinary in the form of contexted unification (Kaplan &
Maxwell). Another would be for programmers to rebel against the
view, so often taken for granted, that this is their problem, and
throw it back to the linguists. This talk will be an intial
exploration of this proposal, based on a well known logic programming
trick attributed originally to Colmerauer and disseminated by
Mellish.
All information concerning this workshop is available at
http://purl.org/net/dm/events/esslli00/
To obtain further information about ESSLLI'2000 please visit
http://events.cs.bham.ac.uk/esslli/
In case of problems or for comments, please contact: dm@sfs.uni-tuebingen.de
(
Last modified: Sun Oct 19 15:25:32 CEST 2008
)