A Generative AI System for Biomedical Data Discovery with Grammar-Based Visualizations
1. This innovative system merges generative AI with grammar-based visualizations to enhance biomedical data discovery. It uses a multi-agent system to generate interactive visualizations and apply filters dynamically, creating a progressive and linked dashboard that adapts to user needs.
2. The system leverages natural language processing to interpret user requests and transform them into visualizations. It supports complex filtering across multiple related data tables, allowing users to refine their data discovery process with both natural language and traditional UI interactions.
3. A key innovation is the fine-tuning of large language models specifically for biomedical metadata visualizations. This ensures that the system can generate accurate and relevant visualizations tailored to the biomedical domain, improving the utility of the interface for researchers.
4. The prototype includes a conversational chat interface and a multi-view visualization panel. Users can send free text requests, and the system responds with visualizations and filter actions, progressively building a comprehensive dashboard that reflects the user’s data exploration journey.
5. The system’s design emphasizes transparency and user control. Actions taken by the AI agents are clearly communicated, and users can adjust or correct the filters and visualizations generated by the system, ensuring a flexible and user-friendly experience.
6. The authors conducted a case study demonstrating the system’s capabilities in a biomedical data discovery scenario. The results highlight the potential of this approach to support complex data exploration tasks and uncover insights that might be missed with traditional interfaces.
📜Paper:
arxiv.org/abs/2509.16454
#GenerativeAI #BiomedicalData #Visualization #DataDiscovery #AIResearch