Constructing Semantic Networks of Comprehension from Eye-movement During Reading

In this research we aim at modeling the interaction between eye-movement and comprehension during reading. Specifically, our work aspires to predict the semantic representation constructed during reading based on eye-movement data. Existing research in psychology has established that cognitive and linguistic processing influence eye-movement during reading. Our work extends existing reading models to generate semantic networks of comprehension based on cognitive theories of eyemovement and comprehension. We evaluate our model through a human experiment in four reading situations explaining and answering many of the phenomena observed in reading research. This work contributes to our understanding of reading as a cognitive task, and provides a framework for computer systems that mimic and evaluate human reading behavior.