Table 2 Prompt variations given to the LLMs
From: Comparative analysis of generative LLMs for labeling entities in clinical notes
Prompt | Prompt |
|---|---|
Variation 1 | “You are an expert in labeling clinical notes with mentions of diseases, symptoms, and medical procedures. Extract the mentions of diseases, symptoms, and medical procedures from the following clinical note: clinical text” |
Variation 2 | “You are an expert in labeling clinical notes with mentions of diseases, symptoms, and medical procedures. Please identify the entities of types [disease, symptom, medical procedure] in the following clinical document. The output must be formatted like this: {diseases: [list of diseases], symptoms: [list of symptoms], medical procedures: [list of medical procedures]} Document: clinical text” |
Variation 3 | “You are an expert in labeling clinical notes with mentions of diseases, symptoms, and medical procedures. Analyze the following clinical note and identify all mentions of diseases, symptoms, and medical procedures. Provide the annotations in a structured JSON format. Each identified entity should be categorized as either a disease, symptom, or medical procedure. Example JSON format for annotations (Note: Replace “entity1”, “entity2”, etc., with actual entity names: { “annotations”: [ { “entity”: “entity1”, “category”: “disease” }, { “entity”: “entity2”, “category”: “symptom” }, { “entity”: “entity3”, “category”: “medical procedure” } ... more entities ] }. Now, analyze this clinical note: clinical text” |