The Research And Application Of Word Embedding Learning Technology In Chinese Sentiment Lexicon Construction

Sentiment analysis is an important branch of the nature language processing.Recently,with the popularity and the development of the Internet,sentiment analysis plays an important role in more and more scenes.For example,sentiment analysis tasks based on product reviews can help companies discover the current market situation of the product which can enable users to take some corresponding strategies for market changes.Moreover,sentiment analysis based on news commentary can be used for public opinion monitoring which can help companies or governments work more specifically.Normally,the emotional expression of subjective texts is mainly reflected by the emotional words and thus the construction of sentiment lexicon is a very crucial basic task which can provide great help for sentiment analysis tasks.For example,sentiment lexicon used as ancillary features can significantly improve the accuracy of sentiment classification tasks.Therefore,we can see that sentiment lexicon has a direct impact on the outcome of the corresponding sentiment analysis task.At present,the method of sentiment lexicon construction based on representation learning can solve the field adaptability problem and have high efficiency and expansion ability.The method uses the word embedding learning model to obtain the key features of words,and then carries out the sentiment classification task of words to obtain the sentiment polarity.The model of word embedding learning plays a crucial role in the process.In this thesis,we propose a word embedding learning model which integrates three kinds of knowledge,including the context words and their composing characters,the polarity of sentences and the polarity of labeled words,based on the existing model and the special features of Chinese language.It is hoped that the word embedding in the construction of Chinese sentiment lexicon can express the semantic and emotional feathers of words better through this method.After that,we apply the model to the construction of Chinese sentiment lexicon in the field of microblog,and a set of practical workflows is given,from which we get a Chinese sentiment lexicon.Finally,the method validation and lexicon evaluation show that the proposed word embedding learning model can improve the effect of the word sentiment classification and applying the model in the construction of Chinese sentiment lexicon is feasible and effective.