MAN: Main-auxiliary network with attentive interactions for review-based recommendation

Recently, more and more attention has been paid to the recommender systems incorporating review information. However, there are two main problems. (1) Among the many reviews written by a user, most of the existing works have not considered the special importance of the user's review for the target item (RT) in building user preferences, which may fail to capture more accurate preferences of the user. (2) Most of the existing work does not dynamically construct the user and the item feature representations in a fine-grained manner according to the aspect characteristics of the target item before user and item nonlinear interaction, which may lead to suboptimal recommendation performance. Therefore, we propose a m ain-a uxiliary n etwork (MAN) based on deep learning for item recommendation. Specifically, MAN uses the auxiliary network to focus on the purification of RT at the word level and assists the main network in generating the predicted value of RT. The main network deals with the user-item interaction according to the relationship between the user multiaspect features and the item as the most prominent aspect feature and then generates the final rating prediction. Note, MAN only uses the main network for testing. Extensive experiments on five public datasets show that MAN outperforms the state-of-the-art methods.