Clean up a raw transcript without changing meaning
Cleans a machine-made transcript by fixing punctuation, fillers and obvious errors, marks every unclear passage and changes no meaning.
Task:
Clean up the raw transcript below so it reads well, without changing what anyone said.
Context:
The recording is [type of recording, for example an interview or a lecture]. The speakers are [speaker names or roles]. The transcript came from an automatic transcription tool.
Rules:
- Fix punctuation, capitalization and paragraph breaks, and remove filler words such as "um" and repeated starts.
- Label each speaker turn with the speaker's name or role.
- Where a word looks wrong, keep it and mark it "[unclear]" with your best guess after it, in the same brackets.
- Do not summarize, reorder or add anything. Do not fix grammar that is part of how the person speaks.
Result:
The cleaned transcript in the order of the recording. After it, a list of at most ten names, terms or numbers I should check against the audio.
Material:
[paste the raw transcript here]