Fact · Speech recognition
5.5percent word error rate
IBM's own English conversational telephone recognizer reached a word error rate of 5.5 percent on the Switchboard subset of the Hub5 2000 evaluation.
Checked against its source on October 17, 2026
In the source's own words
“to the level of 5.5%/10.3% on the Switchboard/CallHome subsets of the Hub5 2000 evaluation”
English Conversational Telephone Speech Recognition by Humans and Machines, George Saon et al., arXiv (cs.CL), IBM Research, 2017, Abstract.
Open the original- What it measures
- The score is for the authors' system on the standard Hub5 2000 telephone test set.
- What it does not show
- A research system measured on a standard benchmark. The authors say the figure was not yet at the level of their human measurements.