- 期刊
Product Feature Sentiment Classification Algorithm based on Bert Model
摘要
Due to the development of the Internet and the comment left people on the Internet, more and more interactive information, by the user the information in the comment on some very valuable, they can help potential customers to understand the product history buyers interested in their characteristics in product reviews the emotional tendency of basic information, such as giving advice, help potential customers to make a decision to buy or not, in addition, these valuable comments information can also help the enterprise operators and managers get consumer demand for product improvement or Suggestions, can help them to improve the quality of the product or service. The purpose of this paper is to explore an efficient feature sentiment classification method for hotel product reviews based on Bert. In this paper, three experiments of Bert-FC, Bert-CNN and Bert-RNN have been carried out. The experimental results show that Bert-FC has the best effect with an accuracy rate of 84.08%, which proves that Bert -FC can be well used in feature sentiment classification.