entity2item: Leveraging Knowledge Graph Embedding for Item Recommendation
Yuan Yuan, Yan Tang, Luomin Du, Xiaotong Li · 2021
Traditional recommender systems only rely on the historical interaction information of users and items, so they often suffer from sparsity problem. Therefore, researchers utilize auxiliary data to alleviate this problem and improve the quality of recommendation. Recently knowledge graph as a kind of auxiliary data has attracted increasing attention, which can extract the semantic information of users and items. However, how to make full use of the information from knowledge graph to build a better recommendation method is still facing numerous challenges. To this end, we propose entity2item, a novel method of using the information from knowledge graphs to assist item recommendation. Specifically, entity2item takes translation-based model as the knowledge graph embedding approach to learn the feature vectors of entities. Then we consider that the items in recommender system are highly correlated with the entities in knowledge graph, so these vectors are introduced to the recommendation module to enrich the expression of items. In the end, the vectors of users and items are input into a deep recommender system for training. Extensive experiments have been conducted on three real-world datasets, the results demonstrate that our method outperforms the state-of-the-art baselines. Even in sparse scenarios, it can still maintain satisfactory performance.