Fact · Languages and accents
0.054mean match error rate
On read non-native English from the L2-ARCTIC corpus, Whisper reached a mean match error rate of 0.054 and AssemblyAI 0.056, which the author describes as approaching human-level accuracy.
Checked against its source on October 16, 2026
In the source's own words
“Results showed that for read speech, Whisper and AssemblyAI achieved the best accuracy with mean Match Error Rates (MER) of 0.054 and 0.056 respectively, approaching human-level accuracy.”
Automatic Speech Recognition for Non-Native English: Accuracy and Disfluency Handling, Michael McGuire, arXiv, 2025, Abstract.
Open the original- What it measures
- Speakers came from six first-language backgrounds (Arabic, Chinese, Hindi, Korean, Spanish and Vietnamese). The read speech was 2,400 single-sentence recordings from 24 speakers.
- What it does not show
- Match error rate is not word error rate. The figure applies to read sentences, not to spontaneous speech, where the best system in the same study had a mean match error rate of 0.063.