Using text mining to glean insights from COVID-19 literature

Area of Science:

• Bibliometrics
• Computational Linguistics
• Infectious Disease Research

Background:

• The rapid increase in COVID-19 publications necessitates efficient information retrieval methods for researchers.
• Existing search methods may not effectively identify emerging research themes within the vast COVID-19 literature.

Purpose of the Study:

• To develop and evaluate a text clustering approach for analyzing COVID-19 research abstracts.
• To demonstrate the utility of text clustering in uncovering latent research topics and streamlining literature review processes.

Main Methods:

• Analysis of 83,264 COVID-19 research article abstracts.
• Application of singular value decomposition (SVD) for dimensionality reduction.
• Utilisation of the expectation-maximisation (EM) algorithm for text clustering.

Main Results:

• Text clustering successfully identified distinct research themes within the COVID-19 literature.
• The clustering approach significantly reduced the volume of literature requiring manual review.
• Emerging and niche research areas were highlighted through the analysis.

Conclusions:

• Text clustering is an effective method for navigating and understanding the large-scale COVID-19 research output.
• This approach aids researchers in efficiently identifying relevant studies and potential areas for future investigation.