I'm training a WORD model with about 50 words. Recognition rate is 100% when using soundfiles from the training with one exception words containing phonemes , and isn't recognised. Don't get me wrong I'm training a WORD model not a phone based. I find this behavior very odd. Of the fifty word seven has or in them and not one of them is regocnised but every other word is. the probability for this to happen by chance is 1 to 100 million .
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Do you use the umlauts in your dictionary or your phone model?
This might be a problem because spinx tries to compare strings case insensitive.
/Dirk
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Anonymous
-
2003-07-02
Chris -- please tell us if you have solved this problem and how. Did it have something to do with upper/lower case of yout phonemes that have diacritical marks?
cheers,
jerry
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Anonymous
-
2003-07-02
Chris -- Sorry, I probably asked the wrong question. In fact, I'm confused about your original message. You are training a word model, so I'm confused about where phonemes are used.
You said that you have seven words containing a-umlaut or o-umlaut. Are those letters in the words or phonemes in the pronunciations?
cheers,
jerry
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I'm training a WORD model with about 50 words. Recognition rate is 100% when using soundfiles from the training with one exception words containing phonemes , and isn't recognised. Don't get me wrong I'm training a WORD model not a phone based. I find this behavior very odd. Of the fifty word seven has or in them and not one of them is regocnised but every other word is. the probability for this to happen by chance is 1 to 100 million .
Do you use the umlauts in your dictionary or your phone model?
This might be a problem because spinx tries to compare strings case insensitive.
/Dirk
Chris -- please tell us if you have solved this problem and how. Did it have something to do with upper/lower case of yout phonemes that have diacritical marks?
cheers,
jerry
Chris -- Sorry, I probably asked the wrong question. In fact, I'm confused about your original message. You are training a word model, so I'm confused about where phonemes are used.
You said that you have seven words containing a-umlaut or o-umlaut. Are those letters in the words or phonemes in the pronunciations?
cheers,
jerry