BioTopicXplor: A Web Tool for Interactive Exploration of PubMed Literature through Reproducible Topics.
Year: 2026
Type: conference-paper
Abstract: The rapid expansion of biomedical literature presents a major challenge for researchers seeking to organize knowledge, identify emerging scientific trends, and explore large PubMed result sets.Existing literature exploration systems primarily rely on predefined ontologies, user-selected features, or clustering approaches that may produce unstable topic structures across runs.We present BioTopicXplor, a tool for interactive exploration of userdefined PubMed article collections through reproducible topic discovery and hierarchical organization.BioTopicXplor is powered by ConvexTopics, an optimization-based topic modeling framework that formulates topic discovery as a convex optimization problem with mixture model parameters drawn directly from the data.Unlike traditional clustering approaches that depend on random initialization and may converge to local optima, ConvexTopics guarantees convergence to a global optimum, eliminates the need to predefine the number of topics, and produces stable, fine-grained, and interpretable topic structures across repeated runs on the same corpus.To support intuitive exploration, BioTopicXplor integrates large language models to generate concise literature-grounded summaries and organize fine-grained subtopics into higher-level thematic groups.The system combines reproducible topic modeling with an interactive human-centric interface that enables users to navigate biomedical literature collections from broad conceptual themes to specific subtopics and associated PubMed articles.
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W7166877683