Research And Application Of The Construction Of Chinese Weibo Emotional Dictionary

In recent years,with the rapid development of Internet technology,the new social media,such as micro-blog and instant messaging tools,have fundamentally changed people’s way of life.The micro-blog based speech information with personal emotion is developing rapidly and in-depth understanding and mining of micro-blog emotional information provide support for government and businesses,etc,institutions’ microblog marketing,brand survey and network public opinion monitoring,and have important social significance and commercial value.In information science,the analysis of emotion refers to use natural language processing and machine learning,etc,to analyze and research the author ’s subjective emotion tendency.Micro-blog emotion analysis is one of the hot issues and there are two main approaches: based on emotion dictionary and based on machine learning.However,in text length,expression style and language style,etc,there is big difference between Chinese micro-blog and traditional text,traditional machine learning cannot keep the relationship between emotional characteristics;at present,the method based on emotion dictionary cannot find available Chinese emotion dictionary with good coverage.Aiming at the shortcoming of traditional research,this paper mainly uses improved SO-PMI algorithm and theme – emotion hybrid model to construct Chinese micro-blog emotion dictionary,which is suitable for Chinese micro-blog emotional analysis,the content mainly involves the following aspects:(1)Aiming at the problem of low coverage of existing emotion dictionary to micro-blog emotional words,to integrate the existing emotion dictionary resources to construct a basic emotion dictionary;at the same time,put forward to improve SO-PMI algorithm by distance mutual information and Laplace smoothing technology,to construct micro-blog emotion dictionary.In addition,the experiment proves that compared with traditional method,in tendency judgment to micro-blog emotional words,the accuracy of the algorithm proposed in this paper has been greatly improved.(2)Study the relationship between the basic emotional words’ emotional tendency and the description of theme in emotion analysis,and propose theme – emotion hybrid model.The model assumes that each micro-blog text in corpus is only consistent with a kind of theme – emotion distribution,and outputs theme – emotion words in the process of model forming document,so as to solve the problem of the same emotional word collocating different themes to show different emotional tendency.Sort out and add theme – emotion words into Chinese micro-blog emotion dictionary.(3)The experiment proves that the effect of using Chinese micro-blog emotion dictionary constructed in this paper to classify micro-blog text emotion is obviously better than the existing emotion dictionary,which verifies the effectiveness of the method of constructing Chinese micro-blog emotion dictionary proposed in this paper.