Towards a Data-Driven Model of Eye Movement Control in Reading

This paper presents a data-driven model of eye movement control in reading that builds on earlier work using machine learning methods to model saccade behavior. We extend previous work by modeling the time course of eye movements, in addition to where the eyes move. In this model, the initiation of eye movements is delayed as a function of on-line processing difficulty, and the decision of where to move the eyes is guided by past reading experience, approximated using machine learning methods. In benchmarking the model against held-out previously unseen data, we show that it can predict gaze durations and skipping probabilities with good accuracy.