An Automated Literature Review Tool (LiteRev) for Streamlining and Accelerating Research Using Natural Language...

Area of Science:

• Computational Biology
• Bioinformatics
• Medical Informatics

Background:

• Literature reviews (LRs) are crucial for synthesizing research but are time-consuming and resource-intensive.
• Traditional systematic reviews face challenges with speed and becoming outdated quickly.

Purpose of the Study:

• To introduce LiteRev, an advanced automation tool designed to assist researchers in conducting literature reviews.
• To evaluate LiteRev's accuracy and efficiency compared to manual literature review processes.

Main Methods:

• LiteRev employs natural language processing (NLP) and machine learning for automated literature searching and topic modeling.
• Techniques include term frequency-inverse document frequency (TF-IDF) matrix representation, dimensionality reduction (PCAM), and clustering (HDBSCAN).
• A k-nearest neighbor (k-NN) search refines results based on user input and selected relevant papers.

Main Results:

• LiteRev processed 631 unique papers, identifying 16 topics and suggesting 193 papers for screening (31.5% of the corpus).
• Achieved a 73.6% recall rate for abstract screening and 87.5% for full-text screening against manual methods.
• Demonstrated a 56% work saved over sampling, significantly accelerating the literature review process.

Conclusions:

• LiteRev effectively streamlines and accelerates literature reviews using NLP and machine learning.
• The tool provides researchers with quick and in-depth overviews of specific research topics.
• LiteRev enhances efficiency and accuracy in synthesizing scientific literature.