Fact · Speech recognition
15.8 to 2.2percent word diarization error rate
In a 2019 Google study, a joint speech recognition and speaker diarization model improved word-level diarization error rate from 15.8 percent to 2.2 percent.
Checked against its source on October 17, 2026
In the source's own words
“our approach improves word-level diarization error rate from 15.8% to 2.2%”
Joint Speech Recognition and Speaker Diarization via Sequence Transduction, Laurent El Shafey et al., Interspeech 2019, 2019, Abstract.
Open the original- What it measures
- The baseline is a conventional system that runs recognition and diarization separately and merges their outputs. The test audio was medical conversations.
- What it does not show
- This is the authors' own experiment on one corpus, not an independent benchmark. The gain may not carry over to other kinds of audio.