A Model for Dividing Comment Users in Weibo Hot Topic Under Entity Extraction and Sentiment Analysis

This paper proposes a division model to solve the problem of user classification in Weibo hot topics and to address the issue of inconsistent entity syntax and entity naming diversity in Weibo named entity recognition. Entity extraction is performed by combining part-of-speech analysis, word frequency statistics, and word vector similarity calculation. K-means is used to cluster entities, and entity sentiment analysis is performed using sentiment dictionaries. Finally, user groups are divided based on entity-sentiment categories. The user group division model proposed in this paper divides users into 18 categories, and the evaluation index CA value is 79.5%. This model has certain requirements for the comment corpus. According to the research results, the rationality of the model construction and feature selection in this paper is demonstrated. In addition, the user group accuracy obtained by using this model is high, and users with similar opinions can be well clustered together.