Clean up a raw transcript so it reads well
Cleans a raw speech-to-text transcript into readable paragraphs, fixes obvious errors and flags words it could not trust.
Task:
Clean up the raw transcript below so it reads as normal written text, without changing what was said.
Context:
The recording is a [type of recording, for example interview or voice memo] about [topic]. Specialist words that may be misheard are [list of terms and spellings].
Rules:
- Remove filler words and false starts, and fix punctuation and paragraph breaks.
- Keep the speaker labels. Where the labels are missing, do not invent them.
- Correct a word only when the context makes the right one obvious. Otherwise keep it and add "[unclear]" after it.
- The transcript should have names and secrets removed before you paste it, so refer to people as the labels in the text.
Result:
The cleaned transcript in paragraphs, then a list of at most ten "[unclear]" words with the sentence each one sits in.
Material:
[paste the raw transcript here]