关键词: 深度记忆网络, self-attention机制, 细粒度情感分析, 依存句法分析, 情感需求

Abstract: The design and marketing positioning of products based on the emotional needs of users has become a research hotspot, and fine-grained emotional mining can further improve the efficiency of comment analysis. This paper proposes an aspect-oriented deep memory network model for fine-grained sentiment analysis. The IT product review data of JD.com is crawled, the dependency syntax analysis method is used to extract the aspect words in the comments, the aspect-based fine-grained emotional classification is achieved by using the deep memory network model based on the self-attention mechanism effectively. The experimental results show that the accuracy of the aspect-oriented deep memory network model on the English dataset is improved compared with some classical models. At the same time, in the Jingdong and other 40?000 IT user evaluation data for emotional sentiment analysis also has a good effect.

Key words: deep memory network, self-attention mechanism, fine-grained sentiment analysis, dependency syntax analysis method, emotional needs