Automated knowledge retrieval from medical texts

Pogodin Ruslan Sergeevich

The article is devoted to the development and application of model theory methods for extracting and formal representing of knowledge from medical documents. The purpose of the research is to automate the filling of the knowledge base with information from texts written in a natural language. The hypothesis is the possibility of applying I. A. Melchuk's theory "Meaning ⇔ text". To extract knowledge, we used methods of converting natural language sentences into quantifier free formulas in the form of fragments of atomic diagrams. Then the formulas were programmatically formalized in machine-readable JSON format. As a result, we implemented a software system interacting with the knowledge base of the IACPaaS cloud platform. For testing we used texts of medical articles describing diseases, their treatment and diagnostics.

Keywords:knowledge retrieval, expert systems, knowledge base, atomic diagram