Peer-reviewed research
Speech recognition error analysis on the English MALACH corpus
Olivier Siohan et al., Interspeech 2004 (ISCA), 2004 · DOI 10.21437/Interspeech.2004-171
Read the originalRead on October 16, 2026
- What it is
- An analysis of word error rate on an English subset of the MALACH oral-history archive, which contains elderly, noisy and accented speech. It finds which recording factors best explain the errors.
- How much weight it carries
- Peer-reviewed conference paper. The recognizer and data are from 2004, and the speech is spontaneous testimony, so the absolute error levels are not comparable with modern engines. The relative findings are still informative.
Facts taken from this source
- The MALACH study observed no evidence that accent or speaker age affected recognition performance.
“while we observed no evidence of the impact of accent and speaker's age on the recognition performance” (Abstract)
- Noise compensation experiments on the MALACH corpus reduced the word error rate by 1.1 percentage points.
“led to a 1.1% absolute reduction of the word error rate” (Abstract)
- In an oral-history archive with noisy, elderly and accented speech, signal-to-noise ratio and syllable rate were the two dominant factors explaining the overall word error rate.
“It was found that the signal-to-noise ratio and syllable rate were two dominant factors in explaining the overall word error rate” (Abstract)
Pages that rest on this source
- Dictation Languages, Accents and Voice TranslationTopic
- How to Dictate Emails Faster Than You Type ThemGuide
- How Accurate Is Speech Recognition? The Real NumbersGuide
- Dictation Mistakes and How to Make Voice Typing AccurateGuide
- Is speech recognition less accurate for some people?Question
- Why doesn't voice recognition understand my accent?Question
- Dictation and Voice Typing: How to Talk Instead of TypeTopic
- Dictation and Background Noise: Where and How to DictateGuide
- How do you improve dictation accuracy?Question
- Noise-canceling microphoneGlossary