Research on Recommendation Algorithm Based on Comment Text

The current recommendation systems based on comment text information generally have the problems of insufficient comment information extraction and low recommendation accuracy. Therefore, this paper proposes a recommendation model integrating global and local features of comment text. Firstly, Convolutional Neural Network(CNN) model is designed to extract global features, secondly, Gating Recurrent Unit (GRU) is designed to extract global features, and finally, Latent Factor Model(LFM) is combined to model users and commodities. Score prediction is carried out by embedding the expression of users and commodities. Experimental results show that the proposed algorithm has good recommendation performance.