Issue with language recognition

[Girish Gopinath]

Hello everyone:

We have a production system that uses Asterisk, UniMRCP and GSR for recognising speech. One of the requirements of the system is to ask the caller to utter the language name for which they require our service. The system recognises the language, converts to text and appropriate action is taken.

This works fine, except for the language Tagalog. When callers say 'Tagalog', the system doesn't find a match. We did some testing and debugging and found that the problem is due to the pronounciation of 'Tagalog'. It is pronounced as "Tuh-gaa-log". Due to this the confidence level is very low and UniMRCP reports a no-match. Below are a couple of results from our tests:

Successful: We pronounced "Taa-gaa-log". Confidence is 0.63

{"transcript":"Tagalog","confidence":0.630923927}],"isFinal":true,"resultEndTime":"7.860s","languageCode":"en-us"}

Failure: Actual pronounciation "Tuh-gaa-log". Confidence 0.27

{"transcript":"Tagalog","confidence":0.276484549}],"isFinal":true,"resultEndTime":"8.040s","languageCode":"en-us"}

We tried modifying the grammar by adding an entry Tuhgaalog, but didn't work. Obviously, we cannot ask the users to say Taa-gaa-log when they call.

Has anyone experienced similar problems? If so, please suggest how you resolved this issue. 

Any help in resolving this is greatly appreciated.

Thank you,

/Girish