Context-aware Dual Representation Learning for Complementary Products Recommendation

Learning product representations that reflect complementary relationship plays a central role in modern recommender system for e-commerce platforms. A notable challenge is that unlike many simple relationships such as similarity, complementariness is often detected from customer purchase activities, which are highly sparse and noisy. Also, standard usage of representation learning emphasizes on only one set of embedding, which is problematic for modelling the asymmetric property of complementariness. We propose using context-aware multi-tasking learning with dual product embedding to solve the above challenges. We encode contextual knowledge into product representation by multi-task learning, in order to alleviate the sparsity issue. By explicitly modelling with user bias terms, we take care of the noise induced by customer-specific preferences. Furthermore, we adopt the dual embedding framework to capture the intrinsic properties of complementariness and provide geometric interpretation motivated by the classic separating hyperplane theory. Finally, we propose a Bayesian network structure that unifies all the components, which also concludes several popular models as special cases. The proposed method compares favourably to state-of-art representation learning and recommendation algorithms for e-commerce, in downstream classification and recommendation tasks. We also develop an implementation that scales efficiently to a dataset with millions of items and customers.