UGRec: Modeling Directed and Undirected Relations for Recommendation

Abstract

Recommender systems, which merely leverage user-item interactions for user preference prediction (such as the collaborative filtering-based ones), often face dramatic performance degradation when the interactions of users or items are insufficient. In recent years, various types of side information have been explored to alleviate this problem. Among them, knowledge graph (KG) has attracted extensive research interests as it can encode users/items and their associated attributes in the graph structure to preserve the relation information. In contrast, less attention has been paid to the item-item co-occurrence information (i.e., co-view), which contains rich item-item similarity information. It provides information from a perspective different from the user/item-attribute graph and is also valuable for the CF recommendation models. In this work, we make an effort to study the potential of integrating both types of side information (i.e., KG and item-item co-occurrence data) for recommendation. To achieve the goal, we propose a unified graph-based recommendation model (UGRec), which integrates the traditional directed relations in KG and the undirected itemitem co-occurrence relations simultaneously. In particular, for a directed relation, we transform the head and tail entities into the corresponding relation space to model their relation; and for an undirected co-occurrence relation, we project head and tail entities into a unique hyperplane in the entity space to minimize their distance. In addition, a head-tail relation-aware attentive mechanism is designed for fine-grained relation modeling. Extensive experiments have been conducted on several publicly accessible datasets to evaluate the proposed model. Results show that our model outperforms several previous state-of-the-art methods and demonstrate the effectiveness of our UGRec model.

MetaKG: Meta-learning on Knowledge Graph for Cold-start Recommendation

“…For example, CKE [52] utilizes TransR [26] to learn item embeddings of knowledge graph. UGRec [55] devises two different embedding spaces to model direct relations from KG and undirect co-occurrence relations among items. However, the embedding-based methods simply adapt the KGE method into recommenders while ignoring the useritem interactions for recommendations.…”

“…• UGRec [55] is the state-of-the-art embedding-based method which not only models directed knowledgeaware relations with TransD [20], but also utilizes attentive mechanism to model undirected co-occurrence relations among items.…”

HAKG: Hierarchy-Aware Knowledge Gated Network for Recommendation

“…• UGRec [55] is a state-of-the-art embedding method that models directed and undirected relations from KG and co-occurrence behavior data. While such undirected relations are inaccessible for other methods and unavailable for the three datasets, we add the connectivities between items which are co-interacted by the same user, and treat them as the co-occurrence relationships.…”

“…• Embedding-based methods [1,4,21,40,42,53,55] directly embed entities and relations in KG via knowledge graph embedding (KGE) methods (e.g., TransR [29] and TransD [22]) to serve as item embedding in recommendation. For example, CKE [53] utilizes TransR to learn item structural representations from knowledge graph, and feeds the learned embeddings to matrix factorization (MF) [35] in an integrated framework.…”