CMOIG · Computational Modeling of Knowledge-Based Inference Generation during Reading Comprehension

The proposed research project employs a pioneering interdisciplinary approach to study inference generation, a fundamental, essential cognitive process that underlies human discourse processing in general and reading comprehension in particular. By integrating linguistic computational modeling (Latent Semantic Analysis and the computational implementation of the Landscape model) with psycholinguistic theories and empirical methods (think-aloud procedures and probing techniques), I intend to explore the extent and types of knowledge-based inferences which are constructed on-line during reading. Although there is a broad consensus regarding the importance of world-knowledge in understanding discourse and achieving text coherence, there is considerable disagreement concerning this issue. Across the various theories of reading comprehension, three central views—Memory-based, Explanation-based, and Coherence-based—can be distinguished based on their conception about the content of the inferences that are activated on-line and about the nature of this process. An original theoretical framework is proposed and examined in an attempt to reconcile the conflicting findings and models under common principles. These principles concern the role of short-term working memory and long-term semantic (i.e., general knowledge) and episodic (i.e., specific text) memory associations between textual and inferred concepts. The theoretical contribution of this research to the understanding of human cognitive mechanisms and its important implications on the development of educational comprehension strategies and the design of intelligence machines that process human natural language are stressed throughout the proposal.