Methods and workflow
Post-editing
Going through and correcting a machine-generated transcript against the audio.
Post-editing is the work of correcting a machine-generated transcript. It has replaced manual transcription as the most common form of transcription work.
The difference from proofreading is decisive: post-editing happens against the audio, not against the text. A transcription error is linguistically correct. It does not fail an ordinary read-through.
What the job consists of
The errors cluster in three categories, and they are not evenly distributed:
- Proper nouns and place names. The single largest item. Surnames are nearly impossible to distinguish by sound.
- Jargon and abbreviations. Internal language, product names, English terms inside local sentences.
- Speaker separation. Especially where several people shared a microphone.
Ordinary words are usually right. Post-editing is therefore targeted reading, not rewriting.
Time spent
Typically 20 to 45 minutes per hour of audio for an ordinary recording, against two to five hours for manual transcription from scratch. With poor audio and many speakers it can double.
How to make it faster
- Listen while you read. Most tools mark where in the audio you are.
- Build a find-and-replace list of the names and jargon your organisation uses. It works for all your recordings.
- Look for repetitions. The same sentence twice in a row is nearly always a hallucination.
- Decide the transcript form before you start. Changing your mind halfway means another pass.