Computational Linguistics Colloquium
Thursday, 12 May 2011, 16:15
Constraining Generalization in Artificial Language Learning: Evidence for rational behaviour in adult and child learners
Language and Cognitive Development Group, University of Oxford
Successful language acquisition involves generalization, but learners must balance this against the acquisition of lexical constraints. For example, certain noun-adjective combinations are restricted in English (e.g. strong winds; high winds; strong breezes; ??high breezes). Another example is the restrictions imposed by verb sub-categorization (e.g. I gave/sent/threw the ball to him; I gave/sent/threw him the ball; I donated/carried/pushed the ball to him; ?? I donated/carried/pushed him the ball). Such lexical exceptions have been considered problematic for acquisition: if learners generalize abstract patterns to new words, how do they learn that certain specific combinations are restricted? (Baker, 1979). Certain researchers have proposed domain-specific procedures (e.g. Pinker, 1989 resolves verb sub-categorization in terms of subtle semantic distinctions). An alternative approach, following Braine 1971, assumes that learners are sensitive to distributional statistics and use this information to make inferences about when generalization is appropriate. This talk presents data from a series of Artificial Language Learning experiments which probed the circumstances in which participants would generalize newly learned linguistic structures. The results demonstrated that, for both adult and child learners, the tendency to generalize is influenced by both lexical frequency and the language-wide statistical properties of the language. I suggest that learners utilize a rational procedure for evaluating and combining difference sources of statistical information.
If you would like to meet with the speaker, please contact