Deep Recommendation Based on Dual Attention Mechanism

Abstract

The recommendation model is committed to filtering massive information and providing users with valuable information. Although the recommendation model based on matrix factorization is widely used, it does not perform well in the case of sparse data. Comments are based on the recommendation model and use comments to extract user preferences and item functions. Although it can alleviate the problem of data sparseness, the interaction between users and items is not a good performance. To this end, we propose a model that can consider the user’s personalized preferences Deep recommendations based on dual attention mechanisms (DRDA). The model is based on the user’s comment text, through the interaction of user comments and project comments, a deep neural network framework with attention factors is obtained to learn the user’s personalized representation. A large number of experiments on the Amazon data set and Yelp data set show that the performance of the DRDA model is better than the traditional baseline model. Further experiments show that the dual attention factor does make a huge contribution to the model.