Mining Eye-Tracking Data for Text Summarization

In this study, we introduce and evaluate a novel extractive text summarization methodology, “SummarEyes,” based on the visual interaction of user with text, using eye-tracking data, as opposed to traditional approaches analysis textual content only. We conducted large-scale study aiming collect data while reading be summarized. utilized various user’s implicit attention metrics generate eye-tracking-based models compared them both typically only single feature gaze duration traditional, well state-of-the-art methods, solely features. The models’ quality was evaluated in terms ROUGE scores intrinsic evaluation datasets had generated, relating behavior personalized DUC gold-standard summaries. experimental results showed that “SummarEyes” significantly outperformed other summarizers predicting generic gold standard With increasing availability technology, research can lead new generation effective user-centric tools.