Quantitative analysis and inference on gaze data using natural language processing techniques
Pascual Martínez-Gómez · 2012
Eye-tracking devices find applications in human-machine interaction, hypothesis testing in psycholinguistic and usability studies, relevant feature extraction when designing models related to human behavior and to build user-centered information systems. We aim at providing a general and robust framework to do quantitative analysis and inference using data collected by eye-trackers when users read text. To achieve this objective, first the accuracy of eye-trackers has to be increased beyond sensor capabilities by using information from the content or the structure of the text. Then, natural language processing techniques will be used to process text appearing on the screen and the recognized reading word sequence. Within this framework, it will be possible to better understand user's intentions, record knowledge acquisition and predict information needs. The intention is to build a user model and user model of the World from texts that users have read. This opens the door to more personalized systems with on-line adaptation capabilities.