Call transcription

Call transcription converts speech from a telephone conversation into written text. It can be produced by people, automated speech recognition or a combination of both. A transcript may include timestamps and speaker labels, depending on how it is created. It is a representation of the audio, rather than a complete replacement for the original conversation or its context.

Accuracy can vary with background noise, overlapping speech, accents, language changes and specialised vocabulary. Speaker separation can also be uncertain, particularly when participants share an audio channel. Names, numbers and important commitments should be checked against the source before the text is used for consequential decisions. Transcription raises privacy and retention questions because searchable text can make sensitive details easier to find and redistribute. Teams should define appropriate access and correction processes. Live transcription and post-call transcription have different timing constraints, so an early live result may change as more of the utterance becomes available.