Smart Searching for the Physiome Project

The Physiome Project's standards for biosimulation models in the Physiome Model Repository (PMR) aim to create a virtual physiological human. Despite an increase in models, users need help finding relevant ones. This thesis proposes approaches to aid search using semantic annotations and Natural Language Processing (NLP). The first approach is NLIMED, which converts free-text queries to SPARQL, while the second one is CASBERT, which encodes model entities for exploratory searches. Integrated into a web-based tool, CASBERT enables comprehensive searches, facilitating model discovery and maximizing FAIRness. These methods can be extended to other standards like SBML, promising multi-scale, multi-repository searches in the future. We further investigate CASBERT to identify unannotated entities by leveraging the hierarchical structure of the model. This exploration led to the development of a functional search tool called BMSE.