Deep Learning-Based Analysis of Emotional Posture in Chinese Literary Works

In order to help readers analyze the emotions covered in Chinese literature and assist them in reading classic Chinese literature.In this paper, we propose a cross-granularity text sentiment analysis model based on interactive attention.Firstly, we carry out the preparation work for text processing of Chinese literary works, and for the shortcomings of the original word vector representation model that lacks the ability to distinguish text weights, we improve the TF-IDF algorithm in terms of weight imbalance of similar corpus, and combine the TF-IDF algorithm with the Word2vec word vector transformation model to enhance the weight of keyword vectors.A cross-granularity text sentiment analysis model based on Interactive Attention Network is proposed, in which the Interactive Attention Network is responsible for receiving the feature vectors encoded by RoBERTa and Word2vec respectively, capturing the bidirectional semantic information of long and difficult sentences.The sentiment analysis model in this paper has different degrees of accuracy and MF1 values on three different datasets.The method is able to provide a better sentiment visualization of some of the story lines of the classic Chinese literary work "Home", and it can accurately capture the emotional relationships in Chinese literature, which helps to assist readers in reading Chinese literary classics.