|
|
Linguistics 682Computational Discourse and Sentence SemanticsMeaning in Context in Natural Language Systems
This course will serve as an introduction
to the use of logical representations for sentences
and discourse, using some of the techniques of computational semantics
to help sort these large issues out. The focus will
be on the interpretation of sentences
in context, in particular, how we are to interpret
natural language expressions
in the context of particular tasks. We will
look at problems involving the semantics of quantification,
plurals, space,
metonymy, and nominal modification and try
to spell out detailed accounts for small fragments
of English, using simplified logical representations.
We will use existing
question-answering, translation, and dialog systems to help sharpen
our questions and to provide examples of approaches
and problems. The course is neither exclusively
a theoretical course nor a computational course,
but rather a course about the confrontation of theory
with practice. Accordingly no programming
knowledge will be assumed.
Practice
The course will begin with Representation and Inference for Natural
Language: A First Course in Computational
Semantics,
by Patrick Blackburn and Johan Bos,
which we will only cover part of.
The purpose of the text is to introduce
to a common logical language and to introduce
techniques for writing semantic rules.
There will be exercises for most
of the text chapters covered, and there will be exercises
outside the book in which students must propose semantic analyses
for phenomena that address certain
practical concerns of a translation or question-answering
system. One recurring issue is that a linguistic
analysis that tries to assign principled meanings
to the parts of sentences seems again and again to fall
short of the richness of interpretations in
context. Thus problem sets and readings
are more of an opportunity to open discussion of
problems than to come up with "right" or
"wrong" answers.
The course begins with an introduction to models and first order
logic, to provide the theoretical basis used throughout. We will then
implement a simple first order logic semantics for a fragment of
English. With the help of
some free software made available with our textbook
(Blackburn and Bos: Representation and Inference for Natural Language)
we will then implement a simple query-answering system.
Having become acquainted with a method, we will then
turn some real semantic phenomena occuring
in corpora and int he semantics literature.
Place and Time
MW 1400-1515 AH2127
Contact Info
Mailing address: |
||||