Text Classification Tool Of Game Community Content Management

With the development of Internet technology and the accumulation of operating experience of game companies,more and more online game communities appear.There are a large number of user generated content(UGC)in the game community.UGC contain both high-quality content and low-quality content.The control of low-quality content is an important part of the game operation.Using machine learning technology to help control the text form UGC can improve the community content operation quality.This paper introduces the application scenario of game community content management,analyzes the problems of existing community content management solutions,and puts forward a scheme of using machine learning to assist game community content management.This paper design and implement a text classification tool used in scheme and verify the network implementation's correctness and effectiveness of the tool including single network Fast Text,Text CNN,upstream and downstream network ELMo,HAN,backbone network BERT.Then text classification tool is applied to the Arena Of Valor and the League of Legends game community to solve UGC text classification,complete dataset construction,model training,and verify that model meet the basic application needs.This paper designs and implements a text classification tool to assist the community content management,which solves the UGC Water Post recognition of the Arena Of Valor game community and text multi classification problems of the League of Legends game community,and improves the community content management ability.