Aggregating large-scale databases for PubMed author name disambiguation

Area of Science:

• Bibliometrics
• Computer Science
• Information Science

Background:

• Author name ambiguity is a persistent challenge in PubMed.
• Existing author name disambiguation (AND) methods for PubMed rely on incomplete or less discriminative internal metadata.
• External databases offer richer, more discriminative information for author disambiguation.

Purpose of the Study:

• To develop a novel author name disambiguation method for PubMed by aggregating information from external databases.
• To enhance PubMed's internal author name metadata using external sources.
• To improve the accuracy and performance of author name disambiguation in bibliographic databases.

Main Methods:

• Aggregated information from Microsoft Academic Graph, Semantic Scholar, and PubMed Knowledge Graph.
• Enhanced PubMed's internal metadata with external and more discriminative data.
• Developed a new disambiguation method, AggAND, integrating enhanced and extended metadata.

Main Results:

• Enhanced name metadata showed performance comparable to existing author identifier systems.
• The AggAND method achieved F1 scores of 95.80% and 93.71% on two datasets.
• AggAND outperformed the state-of-the-art method by a significant margin (3.61% and 6.55%).

Conclusions:

• External databases are crucial for improving author name disambiguation.
• The AggAND method demonstrates the effectiveness of aggregating information from multiple bibliographic databases.
• The proposed methodology can be generalized to other bibliographic databases beyond PubMed.