Using Computational Linguistics to Detect Comprehension Processes in Constructed Responses across Multiple Large Data Sets

Using Computational Linguistics to Detect Comprehension Processes in Constructed Responses across Multiple Large Data Sets Using Computational Linguistics to Detect Comprehension Processes in Constructed Responses across Multiple Large Data Sets This Goal 1, 2-Year Exploration (secondary data) project is an analysis of multiple data sets collected in the context of six previously funded IES and NSF projects in which students produced constructed responses while reading. The proposed analyses go beyond previous analyses by examining the moderating effects of individual differences across multiple tasks (self-explanation, think aloud) and texts. Additionally, the project leverages advanced computational linguistics in concert with dynamical analyses to reveal readers online comprehension processes as properties of text change across sentences. Theories of text comprehension universally assume that comprehension arises with the construction of a mental model for what was read, which requires readers to make connections among information sources (from within and outside the text) and develop a coherent mental model of the text (i.e., coherence building). However, our understanding of how these responses reveal dynamic coherence-building process is incomplete. This project fills this gap by examining properties of constructed responses as they unfold over time and their relations to students individual differences and performance on comprehension tasks.