Contextual-Boosted Deep Neural Collaborative Filtering Model for Interpretable Recommendation

摘要: Abstract Collaborative filtering (CF) is one of the most successful recommendation techniques due to its simplicity and attractive accuracy. However, existing CF methods fail interpret reasons why they recommend a new item. In this paper, we propose Contextual-boosted Deep Neural (CDNC) model for item recommendation, which simultaneously exploits both introductions (textual features) user ratings (collaborative alleviate cold-start problem provide interpretable recommendation. Specifically, an interactive attention mechanism learn representation, makes use mutual information from supervise representation learning each other. With learned weights, can obtain importance historical among list. Meanwhile, assign different weights words in according their importance. Therefore, CDNC interpretations recommendations by assigning historically interacted items introductions. On other hand, also distributed representations new-coming with deep neural networks (i.e., LSTM), considering rating introduction information. Finally, are concatenated perform score prediction. Extensive experiments on four public benchmarks demonstrate effectiveness CDNC. addition, has advantage interpreting providing profiles down-stream applications.