Leveraging Human Reading Behavior for Keyphrase Extraction: A Webcam-based Eye-tracking Corpus

The CLIS-ET corpus integrates webcam-based eye-tracking features, specifically first fixation duration, fixation number, and total fixation duration, into keyphrase extraction models. This method leverages human reading behavior to identify semantically important units in Chinese academic abstracts, addressing the lack of specialized eye-tracking datasets for the LIS domain.