AI Seminar for November 28

[Thomas Bartold]

Our seminar next week will be presented by Rich Thomason, a Professor
in the AI lab. It will be held Tuesday, Nov 28, 4-5pm in the ATL
large conference room. Title and abstract follow.

---------------------------------------------------------------------

Context in Abductive Interpretation

This talk will describe a general approach to contextual
reasoning in natural language interpretation and generation. Drawing
on the view of natural language interpretation as abduction,* I
propose that interpretation provides an explanation of how an
utterance creates a new discourse context in which its interpreted
content is both *true* and prominent.

In explicitly accounting for context in interpretation, the
proposal makes contact with dynamic theories of semantics and
pragmatics, with formal approaches in AI to the representation of
contexts and contextual reasoning, and with computational models of
attentional state in discourse.

I will describe and illustrate a Prolog implementation of
the proposal, and conclude with a general overview of how the
resulting architecture can be deployed.

This is joint work with Matthew Stone, Computer Science,
Rutgers University.