News for hackers. Stuff that matters.

A comment on Researchers achieve speech recognition milestone. Parent comment

Re:Researchers achieve speech recognition milestone by mintplant on Wednesday September 14, 2016 @05:44PM (#12501453)
How does it compare to the typical human error rate?
[ Parent | on HN ]
Re:Researchers achieve speech recognition milestone by M_Grey on Wednesday September 14, 2016 @05:47PM (#12501481)
Best I can find is section four here: http://www.utdallas.edu/~assmann/hcs6367/lippmann97.pdf

It seems to be significantly less than 1% - 1.6%

[ Parent | on HN ]
Re:Researchers achieve speech recognition milestone by pesenti on Wednesday September 14, 2016 @06:02PM (#12501579)
No you looked at the wrong figure. It's figure 7 (Switchboard) and the WER for humans is 4%.
[ Parent | on HN ]
Re:Researchers achieve speech recognition milestone by M_Grey on Wednesday September 14, 2016 @07:32PM (#12502124)
No, if you look that's only the CC component of the test, with and without context; not the whole test.
[ Parent | on HN ]