Table 5 Prompt used for LLM-driven summarization of audio transcripts about COVID-19
From: Generative AI and unstructured audio data for precision public health
Analyze the following transcript and generate a detailed 1-paragraph summary of the speaker’s COVID-19 case that would be relevant to a clinician performing a health assessment. Guidelines: Summarize statements describing experiences related to physical and psychological symptomatology (e.g., ‘I feel short of breath’, ‘I have a fever’, ‘I couldn’t taste anything, ‘my throat is sore’). Do not include any information about names of people, dates (day, month, or year), COVID-19 variants, COVID-19 testing methods (e.g., LFDs, Rapid Tests, PCR tests), COVID-19 medications (e.g., Paxlovid), COVID-19 vaccination status, COVID-19 booster vaccination status, or COVID-19 reinfection history. Do not include the following words and phrases anywhere in the summary: “2020”, “2021”, “2022”, “2023”, “2024”, “omicron”, “delta”,“pcr test”, “rapid test”, “lfd test”, “antigen test”, “paxlovid”, “monoclonal antibodies”, “remdesivir”,“vaccine”, “vaccinated”, “vaccination status”, “unvaccinated”, “booster”, “boosted”,“first dose”, “second dose”, “third dose”, “reinfection”, “reinfected”, “second time”, “multiple infections” If no clinically relevant symptoms are mentioned in the transcript, return an empty string (“”). |
