<div>Dear all,</div><br>Brian MacWhinney will speak about models of
language learning next Wednesday for the second meeting of LTI's seminar
on Computational Linguistics and Natural Language Processing. Details
about the talk are below.<br>
<br>Cheers,<br>Nathan & Ben<br><br>--<br><br>CL+NLP Lunch<b><br>Wednesday, November 3<br>NSH 3305</b><br>Food will be served at 11:45; the talk will begin promptly at noon.<br>
<br><b>Brian MacWhinney</b><br>Professor, CMU Psychology Department<br><br><b>Item-based patterns, computation, and the brain</b><br><br><i>Abstract:</i><br>
Young children build up sentences by combining words into clusters.
Unification grammars such as HPSG, LFG, or Minimalism recognize the
importance of such clusters, but rely on combinations of part of speech
categories whose development is never explained. The alternative
approach to clustering that I have developed emphasizes the role of
item-based patterns in early acquisition. These patterns are initially
specific to individual lexical operators such as “more”, “my” or
“want”. Children then induce higher-level feature-based patterns
through feature pruning, much as in the theory of Hierarchical Bayesian
Models. A left-associative processor can use patterns on these various
levels to generate the required sentence patterns of the target
language.<br>
<br>
In this talk, I will:<br>
<br>
1. review developmental evidence for the shift from item-based to feature-based patterns;<br>
<br>
2. explain how this shift provides a solution to the Logical Problem of Language Acquisition;<br>
<br>
3. examine recent work in computational modeling of language learning
and show why it needs to pay more attention to the shift from item-based
to feature-based patterns; and<br>
<br>
4. link the theory of item-based patterns to core facts about language processing in the brain.<br>
<br><i>Bio:</i><br>Brian MacWhinney is Professor of Psychology, Computational Linguistics,
and Modern Languages at Carnegie Mellon University. He has developed a
model of first and second language processing and acquisition based on
competition between item-based patterns. Data for these models come from
the CHILDES (Child Language Data Exchange System) database, which he
has developed. He is now extending this spoken language database system
to six additional research areas in the form of the TalkBank Project.
MacWhinney’s recent work includes studies of online learning of second
language vocabulary and grammar, neural network modeling of lexical
development, fMRI studies of children with focal brain lesions, and ERP
studies of between-language competition. He is also exploring the role
of grammatical constructions in the marking of perspective shifting and
the construction of mental models in scientific reasoning.<br><br><a href="http://www.cs.cmu.edu/%7Enlp-lunch/" target="_blank">http://www.cs.cmu.edu/~nlp-lunch/</a>