A pipeline that uses a large language model (GPT-4) to standardize free-text clinical notes without altering their clinical content: correcting grammar and spelling, expanding abbreviations and acronyms, normalizing colloquial terms to standard medical terminology, and reorganizing notes into canonical sections. Standardization prepares unstructured notes for downstream concept extraction, ontology mapping, and conversion to interoperable formats such as FHIR, improving readability and data usability while preserving clinical meaning.
This paper presents a large language model pipeline (using GPT-4) for standardizing free-text clinical notes in electronic health records. Standardization here means improving a note's structural and…