講演情報
[6F-03]Boosting Content-based Book Recommendation with Tripartite Matching
*CHEN Chongxian1,2、MO Fan2,1、FAN Xin1,2、山名 早人1,2 (1. Yamana Labratory、2. Waseda University)
発表者区分：学生
論文種別：ロングペーパー
インタラクティブ発表：あり
論文種別：ロングペーパー
インタラクティブ発表：あり
キーワード：
Content-based Recommendation、Book Recommendation、Tripartite Matching
Recommender systems play a crucial role in delivering personalized suggestions, particularly in domains with a vast array of options, such as books. Traditional recommendation approaches often face significant challenges, including the high dimensionality of item embeddings, which results in excessive model parameters and increased training difficulty. In this study, we propose a novel framework based on Tripartite Matching, which introduces a genre vector to mediate interactions between user preferences and item features. By explicitly representing genre information using this vector and separating it from individual item embeddings, our approach significantly reduces embedding redundancy, alleviating the computational burden of training large-scale models. Specifically, we experiment with two Tripartite Matching methods: (1) computing the predicted interaction score as the element-wise product of the user vector, genre vector, and item vector, and (2) calculating the score by summing the pairwise products of these vectors. Experimental results on the Goodreads dataset demonstrate that our approach achieves competitive accuracy compared to state-of-the-art methods while reducing model complexity and improving scalability. This study provides an efficient and flexible solution for enhancing book recommendation systems.