Modelling appraised interest using eye tracking data

Abstract

In this project we use eye tracking data in order to predict implicit user feedback by predicting whether the text was interesting or not. We first study and visualize the data, then apply signal processing methods in order to then extract valuable features and build a classification model. We model appraised interest directly from a set of variables obtained from eye tracking signal. Next, we add to this model predictions of textual complexity and comprehensibility of the text, that are also predicted based on the eye movement features. We show that modelling interest through first predicting textual complexity and comprehension can lead to improved results if carried out properly.