Research And Implementation Of Text Classification For Online Social Platform

Due to the rapid development of the Internet,Natural Language Processing(NLP)has become an important means of acquiring knowledge from the Internet.With the popularity of the Internet,many "mobile phone families" and "thumbs" have emerged,and online social media platforms have become an inseparable part of people's lives.The amount of text data on social media is huge.Through natural language text classification technology,emotional analysis can be focused on the subject areas of interest,so as to make public opinion judgments.The topic mainly solves the topic clustering of social media text content under unsupervised learning and the emotional polarity analysis under supervised learning.Firstly,the topic content of public texts on social media is divided,the topics of research interest are screened,and then the comments data of ordinary users are analyzed under the corresponding themes,and the public opinion tendency is obtained through sentiment analysis.Firstly,a heuristic topic clustering based on Word2Vec is proposed.Through the classification standard of ODP(Open Direc-tory Project),the category center words are set,the word vector is calculated using the Word2Vec model,and then the a priori information of the ODP is used to guide the topic clustering.This can avoid the problem that the classification label of the traditional unsupervised learning method is not clear enough,the classification result is difficult to understand,and the classification method of supervised learning can avoid the requirement of a large amount of annotation data,so that it can be applied to the current social media Shanghai.The content topic classification of the volume data.Secondly,the characteristics of emotional polarity expressed by social network users are different.In view of the shortcomings of current multi-classification methods,a multi-classification model based on integrated learning is proposed.Different types are trained on the labeled data sets.The classification model then uses hierarchical and integrated methods to obtain classification results to improve the accuracy of emotional polarity analysis.