Fact · Speech recognition
4percent word error rate
In the same 1997 review, human transcribers had an error rate of 4 percent on spontaneous telephone conversation, more than ten times lower than the 43 percent of the best machine recognizer.
Checked against its source on October 17, 2026
In the source's own words
“Once again, the 4% error rate of humans is more than an order of magnitude less than the 43% error rate”
Speech recognition by machines and humans, Richard P. Lippmann, Speech Communication, vol. 22, pp. 1-15, 1997, Section 4.4 (Switchboard).
Open the original- What it measures
- The human figure comes from earlier transcription tests the review draws on, not from the same test as the machine figure.
- What it does not show
- Human and machine figures come from different tests. Use the comparison as a rough picture of the gap in the 1990s.
Source
Cited on
- How Accurate Is Speech Recognition? The Real NumbersGuide
- Speech Recognition Benchmarks and LeaderboardsGuide
- Verbatim vs Clean Verbatim TranscriptionGuide
- History of Speech Recognition: From Audrey to WhisperGuide
- Is speech recognition more accurate than humans?Question
- What is transcription QA, and how do you proofread a transcript?Question