Research on Civic Hotline Complaint Text Classification Model Based on word2vec

Abstract

Automatic text classification plays an important role in text mining natural language processing and machine learning. It provides a lot of convenience for information retrieval and text management. In recent years, with the development of Internet technology, text data is rapidly expanding every day, such as microblog dynamic information sent by users, news content of major news portals, e-mail messages from users, posts from forums, blogs, etc. Most of the texts belong to short texts. The short texts have the characteristics of short length, sparse features, and strong context-dependence. Traditional methods have limited accuracy in direct classification. In order to solve this problem, this paper compares the characteristics of various models such as fastText, TextCNN, TextRNN, and RCNN, and the classification effect, trying to find the model with the highest comprehensive ability. Through the use of the Haikou City 12345 hotline complaint text data set for recognition accuracy estimation, the experimental results show that TextCNN has the best classification effect, while fastText has the shortest training time, and TextRNN is not satisfactory in terms of training time or classification effect.