Fact · Languages and accents
46 to 71percent relative word error reduction
In a 2007 study of foreign-accented English, the best combination of pronunciation modelling and acoustic adaptation cut word error rate by 46 to 71 percent relative.
Checked against its source on October 16, 2026
In the source's own words
“The best combination of the above mentioned techniques resulted in a relative word error reduction ranging from 46% to 71%.”
Combined Acoustic and Pronunciation Modelling for Non-Native Speech Recognition, Ghazi Bouselmi et al., arXiv, 2007, Abstract.
Open the original- What it measures
- The experiments used the HIWIRE foreign-accented English speech database and recognizers of 2007.
- What it does not show
- A research result from an older system. A relative reduction of this size does not mean the error rate fell by that many percentage points, and current recognizers may already be far more accurate.
Source
Cited on
- Dictation Languages: What Each App and Device SupportsGuide
- How Speech Recognition Works, From Sound to TextGuide
- Speech Recognition and Accents: Accuracy and FixesGuide
- Dictation for Language Learning: How and Why It WorksGuide
- Can dictation help non-native English speakers write work emails?Question
- Can you train dictation software to your accent?Question
- Can you use voice typing to practice pronunciation?Question
- How do you dictate foreign names and words correctly?Question
- Is speech recognition less accurate in languages other than English?Question
- Acoustic modelGlossary
- Non-native speechGlossary