Glossary
What does “Diarization” mean?
In speech technology, the process of working out “who spoke when” in a recording by splitting the audio into segments and grouping them by voice.
What it means
In speech technology, the process of working out “who spoke when” in a recording by splitting the audio into segments and grouping them by voice. It is how many transcription tools label speakers when everyone was recorded on one audio track.
An example
A recording of a four-person call comes out as “Speaker 1”, “Speaker 2” and so on, because the software grouped the voice segments it found; someone then has to work out which speaker is who.
Not to be confused with: Speaker attribution from separate microphones
Diarization estimates speakers from one mixed recording, so voices that sound alike or talk over each other can be mixed up. When each person’s microphone is recorded separately, the speaker of each line is known rather than estimated.
Where you will see it
In transcription and meeting-notes software, call recording tools, and research on speech recognition.
Questions
Why does diarization get speakers wrong?
Similar voices, people talking over each other, short interjections and poor audio all make it harder to group speech by voice.
Is diarization the same as transcription?
No — transcription turns speech into words; diarization decides who said each part.
Other meeting words
Try Meetful
Meetings that take their own notes
Free video meetings in the browser: every line of the transcript under the name of who said it, and the summary and action items when you hang up. Nothing to install — guests just open the link.