Researchers built an AI tool to sort through 20,766 cancer research papers

Keeping up with new oncology research has become nearly impossible for human experts. A research team from Ukraine addressed this by creating OncoTagger, an automated surveillance tool that scans academic literature to identify relevant AI-related health findings.

By analyzing titles, keywords, and abstracts, the system sorted through nearly 60,000 records indexed in the Web of Science Core Collection. After removing duplicates and filtering, the team narrowed the dataset down to 20,766 specific articles published between 2019 and 2025.

OncoTagger works by applying a predefined set of rules to categorize the studies. In its testing, the tool demonstrated a high level of precision, achieving a 92.3% metric-detection accuracy and a specificity rate of 98.2%.

This system helps researchers visualize the landscape of AI in oncology. It identifies trends in task types, geographic origins of the studies, and potential medical breakthroughs that deserve closer attention.

Could this automated approach help doctors stay current with medical progress without reading thousands of pages every month?

It is important to note that OncoTagger is not a perfect classifier for every individual paper. The creators emphasized that the tool is designed for aggregate surveillance and identifying broad patterns, rather than serving as a definitive census of all AI-oncology research or a replacement for manual peer review.