Text Sentiment Classification Base On Sentiment Series Distance And The Research Of Emotion Subject Recognition Method

Analysis of emotional texts is mainly to find out the user's emotions expressed in the text,analyze the user's attitude.This dissertation studies the evaluation texts on social networks,especially the sentiment texts which are not single or not obvious,and proposes a textual sentiment classification method based on sentiment time series and transition assimilation,and then on this basis,The user's emotions made a more detailed modification of object recognition.The main research is as follows:(1)Combining the sequence of textual emotion and emotional event in the Chinese evaluation system,constructing and extending the dictionary information,quantizing a single emotional sentence into eigenvectors,and then using SVM to construct a classifier function to obtain each The polarity information of emotion-related sentences transforms the whole emotion text into a chronological sequence of emotion nodes,each node representing the polarity of each emotion sentence.After translating emotional texts into emotional time series,we consider the relationship of transitional semantics throughout the textual structure and put forward the concept of transitional assimilation,applying transitional semantics to the level of textual structure.Finally,by calculating the sum of the weighted emotions of each emotion node,the final emotion value obtained is taken as the emotion tendencies of the whole text of the emotion,and the weighting is based on the order in which the affective nodes appear in the emotion timing,the closer to the last of the emotion text Of nodes,have a greater influence on the ultimate emotional orientation of the text.(2)The attitude of the user expressed in an evaluation text may be multifaceted,rather than a single aspect,the opinion expressed by different aspects may be different,so it is identified that the emotional subject of each emotional drama becomes very rich necessary.We first divide the evaluation sentences in the text into non-semantic evaluation sentences and semantic evaluation sentences according to the amount of semantic information in sentences.For the semantic evaluation sentence,because it contains a rich semantic orientation,we use Word2vec to convert each word in the sentence into a semantic vector form,and then sum the weighted vector of all the word vectors in the sentence to obtain the whole semantic evaluation sentence Semantic features form.Then,the semantic evaluation sentence is classified by emotion using the random forest algorithm.In the data set provided in this paper,there are mainly three main emotional subjects.After obtaining the emotional subject of the semantic evaluation sentence,we make the classification of the emotional subject to the non-semantic evaluation sentence by using the rules we set.Finally,to ensure that all the emotional modification subject of the evaluation sentence are identified,the emotion subject contained in the emotional sentence is also identified.Finally,on the evaluation dataset we crawl from a shopping site,we compare the method with different text emotion analysis methods.The experiment results show that the proposed method improves the result of emotional text classification and validates the validity of the method.