Automatic and Controlled Sentence Production: A Computational Model

AbstractWe present a computational model of sentence production that emulates variation of the output of lexicalization and grammatical encoding of the abstract pre-lexical message in terms of complexity and accuracy of the generated sentence as well as fluency and cognitive costs of the sentence production. The model integrates approaches from routine action selection models built on Dual Systems Theory (Norman & Shallice, 1986) with ‘A Blueprint for the Speaker’ developed by Levelt (1989). The paper describes and justifies the model architecture, explores factors affecting language variation in production, and applies the model for testing relationship between complexity, accuracy, and fluency (CAF) of the language production as debated within Second Language Acquisition (SLA) research. The simulation that generated 78750 sentences provides evidence of the trade-off relationship between CAF parameters as speakers have to sacrifice performance on one of the CAF factors in order to improve the remaining two.

Return to previous page