Glossary
Transcription
The written record of what was said in a recording. Accurate transcripts paired with their audio are the raw material for ASR training.
Training-grade transcripts are verbatim: they keep fillers, false starts, and repetitions, and they mark speaker turns and non-speech events. Editorial transcripts, cleaned up for reading, are less useful for ASR because they no longer match the audio. Quality is checked by independently re-transcribing samples and comparing — the same arithmetic as WER.
Why it matters
Audio that ships with accurate transcripts is usable immediately and valued above raw audio alone.