Syntactic priming in corpora as evidence for the sentence production architecture

old_uid3272
titleSyntactic priming in corpora as evidence for the sentence production architecture
start_date2007/10/15
schedule14h30-16h30
onlineno
summaryStructural priming is the tendency to repeat syntactic material in the comprehension or production of language. For instance, speakers have been shown to re-use the structure of a verbal complement after a prime, e.g., either "gave Vera the book" vs. "gave the book to Vera" (e.g., Bock 1986, Pickering and Branigan 1989). I present a series of corpus-based studies that exploit syntactic priming effects as evidence for structural analyses. Using two statistical methods, I demonstrate a short-term and a long-term form of priming. Language production has been shown to vary in its incrementality. On a syntactic level, Combinatory Categorial Grammar (CCG, Steedman 2000) provides a computational, symbolic account of syntax allowing incremental and non-incremental derivations. Syntactic priming can provide us with tests of the flexible constituent structure that CCG predicts. For instance, we can test for structure that varies in its incrementality and, and for combinatory subcategorization and linearization information. Using the notion of "distituents" from the unsupervised-parsing literature, we can show that priming actually acts on structural representations and not just on arbitrary n-grams. The above work has led to a cognitive model of (basic) sentence realization in the ACT-R framework, based on combinatory grammar, which explains priming as a form of learning.
responsiblesInformation non disponible, Crabbé