Eyettention II: A Dual-Sequence Architecture for Modeling Fixation Location, Within-Word Landing Position, and Fixation Duration in Reading
Researchers have developed Eyettention II, a deep-learning model designed to generate realistic eye-tracking scanpaths for reading. This model addresses data scarcity by efficiently training on limited GPU resources and producing complete fixation attributes like location, within-word position, and duration. Eyettention II surpasses current models in scanpath prediction and captures human-like gaze behavior, offering potential advancements in natural language processing and psycholinguistic research. AI
IMPACT This model could enhance NLP applications and psycholinguistic studies by simulating human reading behavior.