Rules for assessing accuracy

[ Follow Ups ] [ Post Followup ] [ Continuous Speech ] [ FAQ ]

Posted by Ross Stuckless on March 01, 1998 at 10:53:27:

I'm engaged in a number of applied research activities pertaining to ASR applications with people who are deaf or hard of hearing, and need a consistent method of assessing transcription accuracy. Developers and users regularly report accuracy by percentage, but rarely if ever report on how they arrived at their figures (or error types). The ARPA standards don't really help much in this regard.

By default I'm working on a protocol now, but would appreciate any comments you might have on the subject, and particularly about any documentation already out there.

Ross Stuckless

Follow Ups:

Re: Rules for assessing accuracy Damien Moody 7/15/98 (0)

Post a Followup


Name    : 

E-Mail  : 



Subject  : Re: Rules for assessing accuracy


Comments:




Optional Link URL:  

Link Title:         

Optional Image URL:

[ Follow Ups ] [ Post Followup ] [ Continuous Speech ] [ FAQ ]

The Sez! Speech Recognition Forum is Sponsored by

Purveyors of the largest selection of speech recognition products worldwide