Construction of a Text Feature Extraction Model Based on Improved SVM for Chinese Elements of Japanese Literature

China and Japan have close ties with each other in various aspects such as economy, politics and culture. Japanese literary works are rich in Chinese cultural elements. It is important to extract texts containing Chinese elements from a large number of Japanese literary works to promote the common development of Chinese and Japanese cultures. Accordingly, the study constructs a Word2Vec-based word vector generation model to extract features of Chinese elements in Japanese literary works, and then constructs an improved support vector machine based text feature classification model to classify the extracted Chinese element features. The experimental results show that when the text vector dimension is 300, the F1 values of TF-IDF-Word2Vec and AVG-Word2Vec are 0.914 and 0.899, respectively, and the F1 value of TF-IDF-Word2Vec method is 0.015 higher than that of AVG-Word2Vec, with significantly better performance than that of AVG-Word2Vec Word2Vec word embedding weighting method. This indicates that the word vector weighting model proposed in the study can extract text features and perform classification relatively accurately. The study provides support for the analysis of Chinese elements in Japanese literary works to promote cultural exchange.