A clustering approach for topic filtering within systematic literature reviews

Area of Science:

• Information Science
• Computer Science
• Bibliometrics

Background:

• Systematic literature reviews (SLRs) involve manually evaluating numerous articles, a process that is time-consuming and lacks transparency.
• Current SLR methodologies present challenges in efficiently filtering large volumes of scientific literature.

Purpose of the Study:

• To develop and evaluate an automated method for filtering and analyzing large article corpora in SLRs.
• To enhance the efficiency, effectiveness, and objectivity of the literature screening process.

Main Methods:

• Application of natural language processing (NLP) on article metadata.
• Utilizing k-means clustering algorithm to group articles into focal topics.
• Analysis of clustering results for topic distribution and scientific community identification.

Main Results:

• Automated clustering significantly improves the efficiency and effectiveness of article filtering compared to manual selection.
• The method successfully converts large article corpora into distributions of focal topics.
• Identified scientific communities and provided an iterative perspective to the linear SLR methodology.

Conclusions:

• NLP and k-means clustering offer a powerful, automated solution for managing large-scale literature reviews.
• The proposed method objectifies the filtering process, making it more transparent and reproducible.
• This approach facilitates quicker identification of research trends and scientific communities within a field.