Fact · Speech recognition
5.9percent word error rate
On the NIST 2000 Switchboard test of telephone conversations, professional human transcribers had a word error rate of 5.9 percent, and 11.3 percent on the harder CallHome conversations.
Checked against its source on October 16, 2026
In the source's own words
“The error rate of professional transcribers is 5.9% for the Switchboard portion of the data, in which newly acquainted pairs of people discuss an assigned topic, and 11.3% for the CallHome portion”
Achieving Human Parity in Conversational Speech Recognition, W. Xiong et al., arXiv (Microsoft Research), 2017, Abstract.
Open the original- What it measures
- Microsoft measured humans and its own system on the same recordings.
- What it does not show
- One benchmark of English telephone speech. Careful transcribers working longer can do better.
Source
Cited on
- Speech Recognition: What It Is and How Well It WorksTopic
- What Is Dictation? Voice Typing and Talk to Text ExplainedGuide
- How Accurate Is Speech Recognition? The Real NumbersGuide
- Human vs AI Transcription: Accuracy, Cost, SpeedGuide
- History of Speech Recognition: From Audrey to WhisperGuide
- History of Dictation: From Shorthand to Speech RecognitionGuide
- How accurate is AI transcription?Question
- Is speech recognition more accurate than humans?Question