Evaluating Co-Authorship Networks in Author Name Disambiguation for Common Names

TL;DR

This paper evaluates the effectiveness of co-authorship network-based methods for author name disambiguation in digital libraries, focusing on common names and proposing community detection to improve accuracy.

Contribution

It introduces an analysis of co-authorship networks for homonym disambiguation and demonstrates how community detection enhances performance for common names.

Findings

Method performs well on most names

Community detection improves accuracy for common names

Optimization needed for homonym disambiguation

Abstract

With the increasing size of digital libraries it has become a challenge to identify author names correctly. The situation becomes more critical when different persons share the same name (homonym problem) or when the names of authors are presented in several different ways (synonym problem). This paper focuses on homonym names in the computer science bibliography DBLP. The goal of this study is to evaluate a method which uses co-authorship networks and analyze the effect of common names on it. For this purpose we clustered the publications of authors with the same name and measured the effectiveness of the method against a gold standard of manually assigned DBLP records. The results show that despite the good performance of implemented method for most names, we should optimize for common names. Hence community detection was employed to optimize the method. Results prove that the applied method improves the performance for these names.