Fact · Voice and AI
4percent word error rate
Training on silent facial EMG reduced the transcription word error rate from 64 percent to 4 percent in one data condition of the 2020 study.
Checked against its source on October 17, 2026
In the source's own words
“decreasing transcription word error rate from 64% to 4% in one data condition and 88% to 68% in another”
Digital Voicing of Silent Speech, David Gaddy and Dan Klein, Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 2020, Abstract.
Open the original- What it measures
- The system turns silently mouthed words, read from facial EMG sensors, into audio, and the audio is then transcribed. The 64 percent figure is the baseline trained only on vocalized speech.
- What it does not show
- Only one of the two data conditions reached 4 percent. The other went from 88 to 68 percent, and the test is in a laboratory with sensors on the face.