I've made a small with 24 words AM in sphinxtrain. When i decode it with s3 accuracy is above 95% while in s4 it lowers drastically to <10%. i tried a whole range of parameters in s4 but no improvement. is there any way to map s3 params to s4 or should i keep experimenting? Or should i try different linguist, frontend or other components with it?
Please help, thank you.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Share your sphinx4 files. sphinx4 accuracy is sometimes even better than sphinx3, but sphinx4 is very sensitive to configuration. So if you did something wrong it will fail.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
hi,
I've made a small with 24 words AM in sphinxtrain. When i decode it with s3 accuracy is above 95% while in s4 it lowers drastically to <10%. i tried a whole range of parameters in s4 but no improvement. is there any way to map s3 params to s4 or should i keep experimenting? Or should i try different linguist, frontend or other components with it?
Please help, thank you.
Share your sphinx4 files. sphinx4 accuracy is sometimes even better than sphinx3, but sphinx4 is very sensitive to configuration. So if you did something wrong it will fail.
thanks again mr. shmyrev. Seems you got solution to all my problems.. well here are the files i prepared ..
http://www.2shared.com/file/2647625/44a8b9d5/s4_data.html