GazeGCN: Gaze-aware Graph Convolutional Network for Text Classification

Bingbing Wang, Bin Liang, Zhixin Bai, Min Yang, Lin Gui, Ruifeng Xu

Abstract: Highlights•The approach takes a fresh perspective by constructing a gaze-aware graph that incorporates gaze signals, emphasizing syntactic information in words and documents.•A novel method leveraging Wasserstein Distance is used to model the relationship between gaze distributions of words, generating gaze-aware word–word weights. Gaze-enhanced TF–IDF is employed to derive gaze-aware word-document weights based on gaze signals.•GazeGCN is proposed to capture deeper syntactic information of words and documents for text classification, achieving state-of-the-art performance on seven text classification datasets.