CAR: Incorporating Filtered Citation Relations for Scientific Article Recommendation

摘要

With the rapid proliferation of information technology, researchers have access to large archives of scientific articles. This makes it more challenging to find articles of interest for researchers. Consequently, a solution to this problem, scientific article recommendation, has become a hot research topic in recent years. In this paper, we propose a novel article recommendation method called Citation-based scientific Article Recommendation (CAR). CAR combines the information of researchers' historical preferences and citation relations between articles. We take into account the fact that, not all pairwise articles with citation relations are highly relevant although researchers generally find articles of interest by searching citations. Therefore, in our proposed method, weak citation relations are first filtered out through an association mining technique using data on researchers' historical preferences. Then, these filtered citation relations are incorporated into a graph-based article ranking method for enhancing recommendation quality. Through a relevant real-world dataset, we evaluate our proposed method. Our experimental results verify that the proposed method significantly outperforms other existing baseline methods in terms of precision, recall, and F1.