Enhancing user identification during reading by applying content-based text analysis to eye-movement patterns

Abstract

In this study, we introduce a brand new idea for a user identification task that benefits from an effective fusion scheme that combines eye movement with syntactic and semantic word relationships in a text. We perform eye-movement recordings during reading because reading process is an instance of high usability as a very common activity. Currently there are very few studies based on eye-movement based identification during reading because of the complex effects of text content on eye-movement behavior. Our proposed method overcomes this drawback by creating a dynamic model for which we register text input and the model’s answer to that input. For this purpose, a vector space representation of text content is interpolated based on fixation duration patterns during reading, leading to high accuracy of identification (an overall accuracy of 98.43%) along with robustness by eliminating the use of common eye-movement characteristics that are sensitive to various factors unrelated to reader identification.