A Deep Learning-Based Framework for the Normalization of WOS Institutional Names

Academic evaluation assesses the contributions and influence of researchers, institutions, or disciplines, determining their reputation and standing within a specific field. As a leading citation database, Web of Science (WOS) is a crucial source for academic evaluation data. However, ambiguities in institutional names due to name changes, multilingual translations, transcription errors, and author preferences significantly impact the accuracy of evaluations. To make data more reliable in academic assessment, this paper addresses deficiencies in deep learning-based techniques for normalizing institutional names in literature. It proposes a framework that combines multiple granular embeddings and multiple context information for synonym detection of WOS institutional names.