Research On Multi-label Emotion Analysis Of Internet Short Text

In recent years,the popularity of Internet technology has deepened,and publishing views and opinions on hot events on the network platform has become an important thing for people to express their opinions.In order to understand the ideas of the general public better,the mass of social networks text resources as an analysis object to monitor and analyze public opinion has gradually become a focus of attention of researchers.At present,many researches on emotion analysis of short texts still remain at the level of sentiment polarity.There is still no unified standard for fine-grained emotion analysis,and single-label classification results often fail to fully reflect the true emotions of users.Therefore,the emotion analysis of short text in social networks under multi-label is a research with very important social significance and value.This thesis is mainly based on the analysis of the content of Twitter,to study the method of identifying the user's emotion from the text and analyzing it.The main work contents and contributions of this article are as follows:First,at present,the multi-label Internet short text emotion tagging corpus is relatively lengthy.In order to better verify the multi-label emotion classification algorithm,a emotion tagging rule for Twitter text is designed.Currently,6,500 tweets and 11,338 sentence emotion corpora have been constructed,in which the twitter level is multi-lab el annotation.Second,for the problem of manual tagging corpus time cost and labor cost is too high,this paper proposes an automated short text emotion tagging method,which completes the annotation of text sentiment polarity and emotion by machine,and on the basis of it The method is revised,which greatly reduces the labor cost of emotional tagging in the corpus.Third,considering that the mainstream text-based emotion analysis methods are mostly simple multi-classification issues,this article conducts multi-label emotion analysis.In the emotion analysis segment,text emotion analysis using the in-sentence features of words for the mainstream alone cannot be accurate to analyze the emotion tendency of Internet short text,this paper improves the MLkNN classifier,combines the adjacent sentences of the sentence and the overall Twitter emotion,and iteratively revises the overall emotion analysis result of Twitter.And on this basis,label correlation is used as one of the conditions for emotion classification,and the analysis results are revised.The improved algorithm has achieved good results in the application of emotion analysis of Internet short text.