Abstract: 【Purpose/significance】With the development of e-commerce, an increasing number of people began to shop online and make comments.These online reviews have become important decision-making information for the business,which is of great signifi⁃cance to understand customer satisfaction and improve product and service quality.【Method/process】This paper combines sentiment analysis and keyword extraction to propose a method for negative keyword extraction of online reviews based on BiGRU-CNN and Tex⁃tRank.On the basis of cleaning more than 60,000 online reviews,the BiGRU-CNN sentiment classification model was constructed to analyze the sentiment of online reviews,and then the keywords of the negative reviews were extracted based on the TextRank method.
【Result/conclusion】The experimental result showed that the method could accurately judge the sentiment orientation of customers' on⁃line reviews.The F1 value reaches 92.41%,and the keywords of negative online review extraction results can better help the merchants to improve product and service quality.【Innovation/limitation】The paper proposes an emotion classification model based on bidirec⁃tional GRU and CNN.On this basis,it extracts keywords from negative comments from emotion analysis based on TextRank method to further improve the accuracy of the model for online comment emotion analysis.