Ranking Scientific Articles over Heterogeneous Academic Network

Due to the explosion of scientific literature, the need for an efficient scientific ranking algorithm has become more important than ever before to assess the importance of scientific articles. The state-of-the-art graph-based algorithms employ the structure of the heterogeneous academic network by mapping the multidimensional relationships between papers, authors and venues into a set of binary relationships. To avoid information loss, this paper proposes a novel mutual ranking algorithm HOMR based on a tensor-based representation of the ternary relationships between academic entities. HOMR is demonstrated effective for ranking scientific publications compared to PageRank, HITS, CoRank and PRank by experiments performed on the dataset and gold standard built on the ACL Anthology Network.