NLM DIR Seminar Schedule

Abstract:

The exponential growth of biomedical literature poses a critical challenge for knowledge discovery, requiring algorithms that can efficiently organize vast information spaces into meaningful structures.
In this talk, we introduce ConvexTopics, a clustering formulation that yields a convex optimization problem, guaranteeing convergence to the global optimum and eliminating the need to pre-specify the number of clusters. Additionally, we integrate ConvexTopics with large language models (e.g., GPT-4o) in a controlled manner to generate literature-grounded summaries and facilitate interactive exploration of computed topics.
As an application, we applied ConvexTopics to more than 12,000 PubMed articles in the domain of anti-aging, uncovering hundreds of coherent and interpretable topics ranging from NAD+ supplementation to Ayurvedic therapies. Beyond longevity science, the methodology is generalizable to diverse biomedical domains, offering a robust framework for large-scale topic discovery, knowledge synthesis, and interactive exploration of scientific literature.