Screening for Reading Deficits using Eye Tracking and Machine Learning

Poster presented at the 2017 Annual International Dyslexia Association Conference. Atlanta, GA.

Abstract:

We present a method that predicts reading deficits in children in just a few minutes with an accuracy of 95 % and a good balance between sensitivity and specificity. We use machine learning of eye movements during reading to find children with reading problems, relative to expectations typical for the age and grade level. Although irregular eye movements are symptomatic rather than causal, our results demonstrate that eye movements can be useful for early discovery of reading deficits. We will show how our method works for English and discuss how it can be used for screening in schools.