Toward a New Paradigm for Author Name Disambiguation

Abstract

Author Name Disambiguation (AND) has emerged as a significant challenge in the bibliometric context with the growing volume of scientific literature. When citations written by different authors have the same names (polysemy or homonym names), and when an author has different names, there is ambiguity (synonyms or name variants). It is difficult to associate a citation with the correct author. Polysemy and synonyms cause merging and splitting anomalies in the citations. These anomalies affect the quantification of an author’s productivity (bibliometric analysis) and the reliability and quality of the information retrieved. Many techniques for AND have been proposed in the literature; most of them do not go beyond string matching or text matching. Most do not consider the context or semantics of the terms used in the citations. The AND problem is resolved semantically in this paper using the deep learning technique on the PubMed dataset. The experimental results show that the proposed method achieves overall (11.72%, 12.5%, and 12.1%) higher precision, recall, and f-measure than the pairwise class classification.