Combining Co-citation and Metadata for Recommending More Related Papers

Abstract

Co-citation is one of the preeminent approach in finding related research articles. The originally proposed technique was solely relying on the bibliographic information but the state of the art in this approach primarily rely on the full-text analysis of articles. The limited availability of full-text limits the applicability of the proposed approaches. Thus, this research used the metadata and bibliographic information of research articles which are openly available. This research explored the hypothesis that traditional co-citation might outperform when combined with metadata relatedness. Similarity scores of different metadata fields (such as title, author and keyword) were calculated and combined with the traditional co-citation relevancy score. The proposed approach has been resiliently tested on the benchmark dataset of 1240 articles of diverse fields of science. The experimental results show that an improvement of 25 percent with co-citation was combined with metadata relevancy score. Further, an interactive visualization has been created to interactively display the resulted documents of co-citation plus metadata analysis.