Fact · Speech recognition
18.5percent word error rate
In a 2012 review, a deep neural network reduced the word error rate of a GMM-HMM baseline from 27.4 to 18.5 percent on a Switchboard telephone task.
Checked against its source on October 17, 2026
In the source's own words
“Using a DNN reduced the word error rate (WER) from the 27.4% of the baseline GMM-HMM (trained with BMMI) to 18.5%”
Deep Neural Networks for Acoustic Modeling in Speech Recognition: The Shared Views of Four Research Groups, Geoffrey Hinton et al., IEEE Signal Processing Magazine, 2012, Switchboard speech recognition task, Table 2.
Open the original (DOI 10.1109/MSP.2012.2205597)- What it measures
- The authors summarize results from the literature in Table 2. This result is on the Switchboard task, tested on the RT03S set.
- What it does not show
- A summary of published systems by four research groups, on one telephone benchmark. Later systems were far more accurate.
Source
Used to calculate
Cited on
- Is Speech to Text AI? What Kind of AI Voice Typing UsesGuide
- Speech to Text Without AI: Dictation With No AI RewriteGuide
- History of Speech Recognition: From Audrey to WhisperGuide
- What are the most important speech recognition research papers?Question
- What role do hidden Markov models play in speech recognition?Question
- Who invented speech recognition?Question
- Acoustic modelGlossary
- Hidden Markov model (HMM)Glossary
- Machine learningGlossary