Abstract In this paper, each text is mapped as a point in a multidimensional vector space in a thought operation model to represent the emotion feature vocabulary in Chinese language literary works in the form of word embedding. The bootstrapping technique is utilized to collect initial seed set vocabulary and target vocabulary, and a corpus of literary works with six basic emotion categories is established. Based on the emotional features of Chinese language literary works, the memory network model for textual emotion analysis based on multiple attention is constructed, and the textual emotion features of Chinese language literary works are empirically analyzed. The results show that the precision is 82.19%, 93.12%, and 94.1%, and the F1Score is 81.93%, 93.4%, and 93.76%, respectively, and the precision and F1Score are better than the effect of the control group model, i.e., it indicates that the effect of sentiment analysis is more excellent. The ratio values of positive emotion, negative emotion, and neutral emotion are, respectively, 63.16%, 15.49%, and 21.35%, i.e., the expression of emotion in Chinese language literary works is more rational. This study provides new perspectives and theoretical references for the study of traditional literary works and has the potential to expand the study in depth.
Read full abstract