Can we specify some words (closely rhyming to KW) as a part of garbage model in sphinx? How to do that?
If we specify such words for detection purpose in the detection list, in addition to keyword, wouldnt it impact the computation as there could be large number of rhyming words to a keyword?
regards
asm
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
My keyword is set to "good day sunshine".
My keyword file has two entries as follows.
"good day sunshine" /1e-20/
"good day" /1/
I am observing that the audios which are spoken as "good day sunshine" are being detected as "good day". Is this normal? Any thing to be done to properly detect the right keyword?
thanks
asm
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am observing that the audios which are spoken as "good day sunshine" are being detected as "good day". Is this normal?
How do you think? If you told it to detect "good day" and it detects "good day". If you don't want to detect "good day" don't add it.
Any thing to be done to properly detect the right keyword?
If you want to detect both "good day" and "good day sunshine" and discriminate between them you might want to modify algorithm to introduce 1 second delay in detection to decide which phrase was actually detected.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am having a multi-accent test data (indian, chinese, arabic, UK and US english). I get high number of errors for indian, chinese and arabic accent when using en-us-semi model. My keyword is "good day sunshine". Is there anything I can do to get a good accuracy on multiple accents? I do not have a large data to do MAP adaptation for all these accents. All I have is a limited test data.
thanks
asm
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am also observing that sometimes, a part keyword is falsely detected as a full keyword. My keyword is "good day sunshine". When I speak "day sunshine", then I detect it as good day sunshine. Is this alright? What can be done to prevent such false alarms?
thanks
asm
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Also a generic question, if I have a four worded keyphrase "ABCD", can i use the keyword file as
A /threshA/
B /threshB/
C /threshC/
D /threshD/
I am noting that after detection of a keyword, all HMMs get cleared and then do transition to some new entry point. Are there any docs which describe the process of clearing and transitioning after keyword detection?
regards
asm
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
Can we specify some words (closely rhyming to KW) as a part of garbage model in sphinx? How to do that?
If we specify such words for detection purpose in the detection list, in addition to keyword, wouldnt it impact the computation as there could be large number of rhyming words to a keyword?
regards
asm
no way to do that
don't add all of them. Try to deal with just a few.
Hi,
My keyword is set to "good day sunshine".
My keyword file has two entries as follows.
"good day sunshine" /1e-20/
"good day" /1/
I am observing that the audios which are spoken as "good day sunshine" are being detected as "good day". Is this normal? Any thing to be done to properly detect the right keyword?
thanks
asm
with respect to my previous post, forgot to mention that I am using en-us-semi model
How do you think? If you told it to detect "good day" and it detects "good day". If you don't want to detect "good day" don't add it.
If you want to detect both "good day" and "good day sunshine" and discriminate between them you might want to modify algorithm to introduce 1 second delay in detection to decide which phrase was actually detected.
hi,
I am having a multi-accent test data (indian, chinese, arabic, UK and US english). I get high number of errors for indian, chinese and arabic accent when using en-us-semi model. My keyword is "good day sunshine". Is there anything I can do to get a good accuracy on multiple accents? I do not have a large data to do MAP adaptation for all these accents. All I have is a limited test data.
thanks
asm
Hi,
I am also observing that sometimes, a part keyword is falsely detected as a full keyword. My keyword is "good day sunshine". When I speak "day sunshine", then I detect it as good day sunshine. Is this alright? What can be done to prevent such false alarms?
thanks
asm
Btw, I've recently committed few important fixes to KWS implementation, now it should work significantly better than before.
thats great !! where can I get it from ?
Also a generic question, if I have a four worded keyphrase "ABCD", can i use the keyword file as
A /threshA/
B /threshB/
C /threshC/
D /threshD/
I am noting that after detection of a keyword, all HMMs get cleared and then do transition to some new entry point. Are there any docs which describe the process of clearing and transitioning after keyword detection?
regards
asm