Neuro-Symbolic Interpretable Collaborative Filtering for Attribute-based Recommendation

Recommender System (RS) is ubiquitous on today’s Internet to provide multifaceted personalized information services. While an enormous success has been made in pushing forward high-accuracy recommendations, the other side of the coin — the recommendation explainability — needs to be better handled for pursuing persuasiveness, especially for the era of deep learning based recommendation. A few research efforts investigate interpretable recommendation from the feature and result levels. Compared with them, model-level explanation, which unfolds the reasoning process of recommendation through transparent models, still remains underexplored and deserves more attention. In this paper, we propose a model-based explainable recommendation approach, i.e., NS-ICF, which stands for Neuro-Symbolic Interpretable Collaborative Filtering. Thanks to the recent advance on neuro-symbolic computation for automatic rule learning, NS-ICF learns interpretable recommendation rules (consisting of user and item attributes) based on neural networks with two innovations: (1) a three-tower architecture tailored for the user and item sides in the RS domain; (2) fusing the powerful personalized representations of users and items to achieve adaptive rule weights and without sacrificing interpretability. Comprehensive experiments on public datasets demonstrate NS-ICF is comparable to state-of-the-art deep recommendation models and is transparent for its unique neuro-symbolic architecture.