AENAR: An aspect-aware explainable neural attentional recommender model for rating predication
Tianwei Zhang, Chuanhou Sun, Zhiyong Cheng, Xiangjun Dong
Abstract
Explainable rating predication becomes challenging with the largely growing number of information and items. Of particular interest is to capture users’ preferences for various items by using textual reviews to achieve accurate and interpretable recommendations. In this paper, we report an aspect-aware explainable neural attentional recommender model for rating predication (AENAR) and this model enables intelligent predication and recommendation by capturing the varying aspect attentions that users pay to different items. The experimental results based on six public datasets reveals that the designed model consistently outperforms five existing state-of-the-art alternatives. Furthermore, the designed attention network allows to highlight the context-aware information in textual reviews that unambiguously suggest users’ aspect-level preference for their desired items, improving the interpretability of the rating prediction.