| Summary: | At present, an increasing amount of research is being conducted on human cognitive behaviour, and reading eye-movement modelling is a research hotspot in cognitive linguistics. However, existing reading eye-movement models are complicated and require a large number of hand-crafted features. To address these issues, this paper improves upon the fixation granularity processing mode and the regression processing mode of the traditional reading eye-movement models and proposes a reading eye-movement fixation sequence labelling method to construct a simpler model. The proposed model is based on a multi-input deep-learning neural network, which takes advantage of deep learning to reduce the number of required hand-crafted features and integrates knowledge from the field of cognitive psychology to increase its accuracy. To meet the data-size requirements of the deep-learning model, this paper also proposes a reading eye-movement data augmentation method. The experimental results show that the proposed method can describe the actual process of reading eye-movement intuitively and that the simple reading eye-movement models based on this method can obtain a similar accuracy with existing models by using fewer hand-crafted features.
|
|---|
