Monday, March 30, 1998
3:00 PM (refreshments 2:45)
Room NE43-518
EECS Special Seminar
Abstract
A grand challenge for artificial intelligence is to understand how children learn language and to get computers to do the same. One interesting and previously unexploited window into the problem of language acquisition are the facts of language change. From one point of view, language acquisition is the mechanism by which language is transmitted from one generation to another. Consequently by studying how languages change and evolve over successive generations, we might be able to infer the mechanisms by which they are acquired by children.
What, then, can we concretely say about the nature of the interaction of acquisition with change? In this talk, I will develop a computational framework to discuss this issue and provide some technical answers.
The facts of language change --- like diachronic change, dialect formation, language evolution, lexical diffusion and so on have received minimal computational attention in the past. I will argue that historical phenomena are typically at the group level. They require us to understand the behavior of a population of linguistic agents. I will show how to derive such population behavior starting from an understanding of the behavior of the individual. By modelling a learner and aggregating over a population of learners, one can formally derive dynamical systems models of language change. I will apply these models to the cases of Portuguese, English, and French. I will comment on several extensions of the basic model and point to similarities between this approach and that taken in population biology and evolutionary economics.
|
Modified: Mar 16, 1998|
Current events|
Your comments
and inquiries are welcome.