Recurrent knowledge graph embedding for effective recommendation

TLDR

Knowledge graphs have proven effective for recommendation, yet existing methods rely on hand‑engineered features such as meta paths that require domain knowledge. This paper introduces RKGE, a KG embedding approach that automatically learns semantic representations of entities and the paths between them to characterize user preferences for items. RKGE uses a novel recurrent network architecture comprising a batch of recurrent networks to model path semantics between entity pairs, fusing them into recommendations, and applies a pooling operator to weight the saliency of different paths. Validation on real‑world datasets demonstrates that RKGE outperforms state‑of‑the‑art methods and offers meaningful explanations for its recommendation results.

Abstract

Knowledge graphs (KGs) have proven to be effective to improve recommendation. Existing methods mainly rely on hand-engineered features from KGs (e.g., meta paths), which requires domain knowledge. This paper presents RKGE, a KG embedding approach that automatically learns semantic representations of both entities and paths between entities for characterizing user preferences towards items. Specifically, RKGE employs a novel recurrent network architecture that contains a batch of recurrent networks to model the semantics of paths linking a same entity pair, which are seamlessly fused into recommendation. It further employs a pooling operator to discriminate the saliency of different paths in characterizing user preferences towards items. Extensive validation on real-world datasets shows the superiority of RKGE against state-of-the-art methods. Furthermore, we show that RKGE provides meaningful explanations for recommendation results.