Sentiment Analysis of E-commerce Text Reviews Based on Sentiment Dictionary

With the rapid development of the Internet, the advantages of e-commerce in marketing have become prominent. However, it is difficult for consumers to choose a variety of products of the same type. Consumers give feedback on the purchase process in the form of comments, and in this way influence other users' purchase decisions. This paper mainly divides the text comments of e-commerce from the three aspects of keywords, evaluation objects and emotional resources, and classifies them according to the constructed emotional dictionary. Specific experimental steps: Firstly, the tf - idf algorithm to extract of subjects of keywords. Secondly, by using the method of part of speech combining keywords similarity evaluation object extraction, and based on the characteristics of parts of speech and location rules set to extraction of emotional resources. Finally, based on the emotional resources to build a dictionary, and for the same emotional words to different evaluation objects with different polarity, constructing reverse dictionary. The experimental results verify that the emotion dictionary constructed in this paper has a good effect on the emotion classification of e-commerce text reviews.