Using Elasticsearch for entity recognition in affiliation disambiguation

TL;DR

This paper presents a modular Elasticsearch-based method for automatic affiliation recognition in scholarly metadata, enabling customizable precision and recall in aligning with registries like countries, GRID.ac, and RNSR.

Contribution

It introduces a flexible, user-controlled alignment approach using Elasticsearch for affiliation disambiguation in scholarly data.

Findings

Effective alignment on three registries demonstrated

Method allows user-defined precision and recall

Open-source implementation available on Github

Abstract

Automatic recognition of affiliations in the metadata of scholarly publications is a key point for monitoring and analyzing trends in scientific production, especially in an open science context. We propose an automatic alignment method on registries, based on Elasticsearch. The proposed method is modular and leaves the choice of the alignment criteria to the user, allowing him to keep control over the precision and recall of the method. An implementation is proposed for an automatic alignment on three registries: countries, GRID.ac and RNSR (research laboratory directory in France) on the Github https://github.com/dataesr/matcher and the performances are analyzed in this paper.