Speech recognition accurary

[sanil]

Hi Brian :
This library is a big help. I am working on an iphone app to convert
speech to text and reverse and am testing your library. Currently, I
am playing with VocalKitTest and it does a great job of converting
text to speech but it seems to have a hard time converting my speech
to text with accuracy. Any pointers on what I should be looking for ?
I will be playing around with the pocketsphinx.conf settings but am
not sure if that is the problem.

~sanil

[Brian King]

Hey Sanil,
Accuracy with open ended dictation is pretty poor. If you
restrict the word dictionary, you can get better results. There's a
page on the wiki, but I think there's another better page that I can't
find the link to. Anyone out there have the link?

There's also a chance that your accent can be affecting results. I
wouldn't go down that route though until you have looked at
restricting the dictionary. I wouldn't lean on pocketsphinx/VocalKit
to do open ended dictation.

> --
> You received this message because you are subscribed to the Google Groups "VocalKit" group.
> To post to this group, send email to voca...@googlegroups.com.
> To unsubscribe from this group, send email to vocalkit+u...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/vocalkit?hl=en.
>
>

[Sanil Pillai]

yeah, that's true. I am using the lmtool at cmusphinx to create a restricted dictionary and having much better results.  This is the link -> http://www.speech.cs.cmu.edu/tools/lmtool-new.html

 

Thanks.

 

~sanil