Visualizing Article Similarities via Sparsified Article Network and Map Projection for Systematic Reviews.

Area of Science:

• Biomedical Informatics
• Information Science
• Data Visualization

Background:

• Systematic reviews (SRs) synthesize biomedical literature, a process that is labor-intensive due to large article collections.
• Identifying relevant articles efficiently is crucial for SRs, and textual similarity analysis has shown promise.
• Current methods for visualizing article relationships may lack interpretability or scalability for practical SR use.

Purpose of the Study:

• To develop and evaluate novel visualization techniques for article similarities in SRs.
• To enhance researchers' comprehension of article distributions and uncover hidden patterns within literature collections.
• To extend the practical utility of article similarity analysis for efficient SR article screening.

Main Methods:

• Implemented a graph-based network visualization approach.
• Utilized three distinct network sparsification techniques to manage complexity.
• Employed dimensionality reduction for distance-based map projection visualization.
• Compared the effectiveness of network visualization against map projection.

Main Results:

• Demonstrated the ability of visualizations to reveal article distribution patterns.
• Showcased the identification of meaningful clusters of relevant articles.
• Evaluated and compared the performance of different sparsification methods and visualization types.
• Confirmed the utility of the developed methods in practical SR settings.

Conclusions:

• Visualization of article similarities offers a powerful tool for SR researchers.
• The implemented methods improve comprehension of literature landscapes and aid in identifying relevant article clusters.
• This approach can significantly enhance the efficiency and effectiveness of the article screening process in systematic reviews.