Sentiment Classification of Internet Commodity Reviews Based on the Extended Chinese Sentiment Lexicon

Abstract

The internet commodity review is the purchaser’s feedback on the goods. Massive commodity reviews contain the purchaser's sentiment orientation, which has important research values for enterprises. Meanwhile, how to identify sentiment orientation from these reviews is also a hot spot in the field of natural language processing. In this paper, we presents a method of sentiment lexicon extension based on word2vec, to construct a sentiment lexicon suitable for the field of e-commerce. In addition, we improved the judgment rules of sentiment classification in consideration of context. The experiment results shows the model in this paper has improved about 14 percentage points in recall rate, accuracy rate and F value compared with the traditional model. Therefore, the effectiveness and accuracy of the model is validated.