Research evaluation metrics comparison: does citation or citation context count differ from citation context similarity metric?

The paper introduces a citation weighting method that leverages semantic similarity scores to assign greater weight to unique citation contexts, while minimizing the influence of repeated ones in scientific publications. The study sample consisted of 100 highly cited biomedical publications and their citations. The 7317 citations that were accessible generated 11,228 citation contexts and 9,795 citation context pairs. Semantic similarity scores between citation context pairs were obtained using BioSent2Vec word-embedding model for the biomedical publications. Weights were assigned to the cited document based on the uniqueness of its citation contexts in the citing document. The proposed method was evaluated against the existing metrics to assess differences, utilizing case studies, rank change analysis, and statistical methods. Results of the Spearman’s rank correlation tests showed that the citation context similarity metric correlated highly with citation count and citation context count. However, the case study and rank change analysis indicate that the proposed method ranks publications differently from either the citation context count or traditional citation-based metrics. The rank change analysis and case studies showed that the proposed citation context similarity was different from citation context and citation counts because of the proposed method’s context awareness, which enables it to assign weights based on the uniqueness of citation contexts. Although the proposed citation weighting scheme involves more complex computational requirements, it is recommended as a credible and viable alternative to the traditional citation context count.