Research on Matrix Multiplication Based WeChat Group Tagging Technology

Abstract

WeChat is one of the most popular instant messaging tools in the world. WeChat group is a group of more than two users with several common characteristics or goals. Tagging WeChat group is of great significance for WeChat group management and the follow-up WeChat group portraits, WeChat user portraits, and other academic researches. This paper regards tagging as a classification problem for WeChat group names. Firstly, we preprocess the group name with four rules. Then, we use machine learning classification methods to classify WeChat group names. Traditional machine learning method could achieve 95% accuracy, but the memory and time costs are relatively large. As a result, it is not suitable for tagging millions of WeChat groups. In the paper, we propose a method based on matrix multiplication to tag WeChat group names, which can reduce the memory and time consumption in model training phrase with the circumstance of maintaining almost the same accuracy. Our evaluation shows that it can complete millions of WeChat group name tagging tasks properly.