A scholars’ personality traits augmented multi-dimensional feature fusion scholarly journal recommendation model

Journal recommendation is a popular research topic in academic resource recommendation. However, the reliability of the current model depends on rich features in the dataset, and ignores the issue of model performance being degraded by sparse sample features. To tackle this issue, inspired by personality trait-based recommendation techniques, we propose a Personality Trait-augmented Multi-dimensional Feature Fusion Journal Recommendation (PTMFFJRec) model that integrates scholars' personality traits and multi-dimensional deep semantics, and utilize linguistic features and the big-5 personality model to estimate the personality of the scholars. This is the first multi-dimensional feature model that incorporates Transfer Learning, BERT, and GCN techniques to recommend academic journals based solely on the abstracts and titles of submitted manuscripts. Experimental results on the real-world Scopus's dataset demonstrate that PTMFFJRec outperforms advanced benchmark models, specifically, surpassing the baseline models in metrics of MAP, MRR, Recall@20 and Diversity.