【摘要】 将知识图谱作为外部信息引入推荐系统可以有效缓解推荐系统的数据稀疏问题。文章提出一种端到端的神经网络模型，文章使用一种图注意力机制取代基于相似度或交换矩阵计算的离线元路径方法；另外，根据知识图谱中的项目的不同实体类型，文章提出一种多视图记忆注意力网络去学习更深层次的项目表征。文章在MovieLens数据集上进行了实验，实验结果表明，本文的模型明显优于Top-N推荐基线模型。更多还原

【Abstract】 Exploiting the external information called knowledge graph to recommendation has shown to effectively alleviate the data sparsity problem of recommendation system. Article proposes an end-to-end neural network model. We use a graph attention mechanism to replace the offline meta-path method based on similarity or commuting exchange matrix. In addition, according to the different entity types of items in the knowledge graph, we propose a multi-view memory attention network to learn more profound item features. Experiments on MovieLens dataset show the effectiveness of our model significantly outperform baseline model for Top-N recommendation.更多还原