AI: Mark Steedman University of Edinburgh The Computational Problem of Natural Language Acquisition ACES 2.402 Wednesday March 7 2007 11:00 a.m.

Contact Name: 
Jenna Whitney
Mar 7, 2007 11:00am - 12:00pm

There is a signup schedule for this event.

Type of Talk:

Speaker/Affiliation: Mark Steedman Professor University of E


Date/Time: Wednesday March 7 2007 11:00 a.m.


ation: ACES 2.402

Talk Title: The Computational Problem of Natural
Language Acquisition

Talk Abstract:
The talk reviews work-in-pro

gress on language acquisition in children and robots using combinatory cate

gorial grammar (CCG) building on work by Siskind Villavicencio and Zettl

emoyer among others.

CCG is a theory of grammar in which all langua

ge-specific grammatical information resides in the lexicon. A small univers

al set of strictly type-driven non-structure dependent syntactic rules (b

ased on Curry''s combinators B S and T) then projects lexical items into

sentence-meaning pairs. The task that faces the child in the earliest stage

s of language acquisition can therefore be seen as learning a lexicon on th

e basis of exposure to (probably ambiguous possibly somewhat noisy) senten

ce-meaning pairs given this universal combinatory projection principle an

d a mapping from semantic types to the set of all universally available lex

ical syntactic types.

The talk argues that a very simple statistical
model allows children to arrive at a target lexicon without navigation of

subset principles or attention to any attendant notion of trigger other th

an the notion reasonably short sentence in a reasonably understandable situ

ation. The model explains the pattern of errors that have been found in eli

citation experiments. The linguistic notion of parameter appears to be redu

ndant to this process.

The talk goes on to consider some more genera

l implications of the theory including its application to the phenomenon o

f syntactic bootstrapping touching on the question of the prelinguistic or

igin of the combinatory projection principle itself.

Speaker Bio:
Mark Steedman is Professor in the School of Informatics at the

ty of Edinburgh. He received his PhD from the
University of Edinburgh in
1973. He came to Edinburgh in 1998
from the University of Pennsylvania
where he was Professor in
the Department of Computer and Information Sc

ience. He is a
Fellow of the American Association for Artificial Intelli

a Fellow of the Royal Society of Edinburgh and a Fellow of theBritish Academy.

His research interests cover issues in computatio

linguistics artificial intelligence computer science and

tive science and their applications in practical systems
including syn

tax and semantics of natural language
wide-coverage parsing comprehens

ion of natural language by
humans and by machine and the role of intona

tion in spoken
language generation and analysis. Some of his research co

the analysis of music by humans and machines. He has acted as

advisor for twenty-four PhDs.