Network Neologism Recognition And Polarity Analysis Based On Explicit Semantic Analysis

With the rapid development of the Internet era,the use of short text forms such as micro-blog and WeChat has increased dramatically.Especially in the era of data explosion,microblog has become the biggest platform for the formation and spread of new words on the Internet.In the field of natural language processing,the inaccurate recognition of these Internet neologisms leads to difficulties in the analysis and understanding of such texts,especially in the recognition of network neologisms and their semantic and emotional understanding.Lack of the ability to interpret new words means that users can not understand the content of the expression,nor can they conduct the correct guidance of public opinion.Therefore,it is of great significance for the identification,semantic understanding and emotional analysis of network new words.This paper regards the discovery of network neologisms and the analysis of their emotional polarity in micro-blog as the research topic.The specific research contents are as follows:In short text sentiment analysis,the emotional polarity of words is the core of judging short text emotion.New words are widely used by Internet users because of their simple and interesting expressive ability.However,there are no such new words in traditional corpus,which makes it difficult to judge the emotional polarity of new words,thus affecting the final results of short text emotional analysis.The primary requirement of completing the above research work is the accuracy of word segmentation.The accuracy of word segmentation results directly affects the accuracy of emotional analysis results.Network neologisms in unknown words are the main factors affecting the accuracy of word segmentation.In order to overcome the shortcomings of traditional methods that can not recognize network neologisms,a new network neologisms recognition method based on Explicit Semantic Analysis(ESA)and network entity links is proposed.The purpose of this method is to retain the logic of the original text as much as possible to prevent misreading.After word segmentation,the existing corpus is used for semantic analysis,and the unlisted words are re-analyzed by ESA method to get the final semantic results.Based on this research,a new method of emotional orientation recognition is proposed,which combines the mutual information of word vectors and emotional points.This method mainly overcomes the difficulties of emotional orientation recognition of network neologisms and expression symbols.The experimental results show that,compared with the existing new word recognition algorithms,this algorithm only needs a small amount of corpus as the underlying knowledge support,which greatly reduces the cost of artificial rules formulation,and improves the accuracy of network new word recognition and emotional understanding.