Dr Mercedes Arguello Casteleiro: Abstract

Digitalisation is happening in both human and veterinary medicine, however current Artificial Intelligence (AI) "cannot reliably deal with facts, perform complex reasoning, or explain its conclusions". Hospitals produce petabytes of data per year, most of which goes unused. Most of healthcare data (estimated 80%) is unstructured, which can be made more meaningful by classifying the data and normalising the data with internationally recognised standards, such as SNOMED CT as clinical coding standard.

Transforming the ever-changing body of scientifically sound research (evidence) from human-readable to machine-interpretable requires both biomedical/clinical knowledge and reasoning capabilities. This talk will present AI approaches for transforming a body of evidence into knowledge graphs (symbolic AI) supporting question answering from junior doctors. The results of a usability testing with 13 UK junior doctors will be presented. Come along if you also want to hear about a Neuro-Symbolic AI approach with biomedical/clinical and general Large Language Models (LLMs from neural AI) that aims to lower the technical skills overhead (understanding of AI and programming code) by leveraging on free Web-based tools (Low-Code/No-Code).

Can small open-source biomedical/clinical LLMs compete with free-of-charge general LLMs (DeepSeek, Groq, Grok3-beta, QWEN2.5-MAX, Gemini, Claude, ChatGPT4-o)? to answer this question, a natural language processing task was proposed.

Organised by BCS SGAI The Specialist Group on Artificial Intelligence http://www.bcs-sgai.org