Skip to content

Fact · Speech recognition

3.5percentage points

On CallHome, the IBM recognizer's word error rate was 3.5 percentage points higher than the best human transcriber's.

Checked against its source on October 17, 2026

What it measures: Both numbers come from the same 2017 study and the same test set.
How it is calculated: 10.3 - 6.8 = 3.5
What it does not show: The comparison uses the best human transcriber, not the average one, and a single test set.

Calculated from

IBM's own English conversational telephone recognizer reached a word error rate of 10.3 percent on the CallHome subset of the Hub5 2000 evaluation.
In the same IBM study, the best human transcriber reached a word error rate of 6.8 percent on the CallHome telephone test.

Cited on

History of Speech Recognition: From Audrey to WhisperGuide

CallHome gap between machine and human: 3.5 points | Yap