A journal name semantic augmented multi-dimensional feature fusion model for scholarly journal recommendation

Abstract

Recommending appropriate academic journal to researchers has become a time-consuming and challenging task. In this paper, we propose a Journal Name Semantic-enhanced Multidimensional Feature Fusion Journal Recommendation (JNSMFFJRec) model that both integrates journal name semantic information and capture deep semantic information. The model is the first of its kind to fuse multi-dimensional feature based on BERT layer and tow-tower layer, that only requires the abstract and the title of a manuscript to identify academic journal. Compared to advanced benchmark models, the experiments based on the real dataset Information System and Management (ISaM) of Scopus show that the performance of the proposed academic journal recommendation model outperforms. Specifically, the Mean Average Precision improved by 1.34%-19.42%; the Mean Reciprocal Rank improved by 3.56%-59.79%; the TOP-5 Recall improved by 1.72%-43.94%; and the result Diversity improved by 2.4%-18.6%.