Research And Implementation Of Group Chat Public Opinion Monitoring Technology Based On Text Mining

There are many data types of group chat messages,and the feature vector of text content presents the characteristics of high dimension and sparse,so the mainstream clustering methods are difficult to achieve good results;different group chat topics are easy to cross and parallel,and the existing technology of segmentation and extraction of topics still has a large space to improve;in the aspect of group chat public opinion monitoring,how to improve the accuracy of sensitive content identification and ensure the timeliness of early warning information is an important issue.To solve the above three main problems,we need to use computer technology to carry out public opinion monitoring on group chat data.In this paper,based on the characteristics of group chat and text mining technology,for the purpose of realizing public opinion monitoring,the concept of group chat topic sequence is proposed.The topic sequence is designed and constructed based on topic attributes such as the frequency and duration of the topic recently discussed,and the popularity of group chat topics generated or updated in real time can be effectively judged.A multi-strategy-based group chat topic detection algorithm is proposed,which combines message attributes such as sender,sending time,message type and other message attributes to improve the clustering effect of streaming data.Combined with topic sequence,it solves the problem of group chat topic cross-parallel.This greatly improves the performance of group chat topic dynamic extraction.Finally,real-time monitoring is carried out for the subject content of group chat topics,and public opinion warnings are made for sensitive content in new messages,and the analysis results are persistently stored in the database,and presented to the front-end page of the public opinion monitoring system at the same time.The experiment of this subject first uses software reverse technology to collect real-time group chat messages received by the PC version of the WeChat client.Compared with the best algorithm in recent years,the group chat topic dynamic extraction algorithm proposed in this article has an average F value of 2.9%,6.0%and 2.0%higher on the three sample data sets,and the speed has been greatly improved,they are 27.7%,32.1%and 47.1%higher respectively.The early warning model used in this topic can also more accurately detect sensitive topics in group chat records,and establish associations between historical topics and the latest news through sensitive keywords,so as to facilitate the monitoring of topic evolution trends and improve group chat public opinion early warning accuracy.