Indexado en

Licencia y uso

Citaciones

Altmetrics

Grant support

This work was partly supported by Grant No. PID2022-136374NB-C22 funded by Ministerio de Ciencia, Innovacion y Universidades and Agencia Estatal de Investigacion (Spain), by the Aragon Government through the research group E30_23R and by the Universidad de Zaragoza under the temporary research contract program "Programa Investigo" (Programa Investigo-081-74), funded by the Servicio Publico de Empleo Estatal and the European Union-NextGenerationEU.

Análisis de autorías institucional

Compartir

3SA: an entity-linking algorithm for the Institution Name Disambiguation problem in affiliations using edit distance

Publicado en:Scientometrics. 130 (7): 4073-4091 - 2025-07-16 130(7), DOI: 10.1007/s11192-025-05368-1

Autores: Muñoz-Jordán D; Ruiz G; Cabriada P; Durán JL; Iñiguez D; Rivero A

Afiliaciones

Resumen

When researchers sign an article, they reference all the institutions they belong to, writing one or more affiliations containing them. Researchers sign in many different ways, and different journals also have varying standards in this regard. In this article we will focus on the Institution Name Disambiguation (IND) problem, also known as Organization Name Disambiguation (OND). Common issues associated to IND problem arise because researchers may write the name of the institution differently in various publications, and different researchers from the same institution will certainly write it differently as well. On the other hand, a researcher may be affiliated with several centers simultaneously or at different stages of their professional life, which introduces the factor of time as an additional variable to consider. As a result, analyzing and linking scientific work from different areas for various institutions is challenging. Databases like Web of Science collect articles from various journals across different fields. In this article, we will propose a method named 3 Steps Affiliation (3SA) based on, firstly, preprocessing the information, secondly, candidate extraction via localization and classification type of the institutions and, thirdly, on entity linking to extract the institutions from affiliations downloaded from Web of Science articles using an edit distance. We use a world-wide open source database with more than 100k institutions to solve the Institution Name Disambiguation problem. We show that the proposed method has a state-of-art performance by comparing it with other methods. Additionally, we evaluate the impact of different edit distance metrics within our method to identify which yields the best results.

Palabras clave