Please provide the following information when requesting support.
Hardware - GPU (3090)
Hardware - CPU (Intel i9)
Operating System (Ubuntu 22.04)
Riva Version (v1.9)
TLT Version (–)
How to reproduce the issue ? (This is for errors. Please share the command and the detailed log here)
I am attempting to boost certain words/phrases by setting a SpeechContext in the RecognitionConfig of a RecognizeRequest. The boost seems to work, but Riva seems to permanently set those words as boosted.
In other words, if I make a RecognizeRequest with some boosted words, followed by a new RecognizeRequest without any boosted words, the second request is still using the boosted words from the first request. The only way I’ve found to get rid of them is the “deboost” them (send another RecognizeRequest including the boosted words but with a negative boost value).
I was wondering if this is the expected behavior. I’m guessing that what is really going on is that when you add boosted words, they get permanently put into the lexicon, and the only way to remove them is to either “de-boost” them or restart Riva.
Sure would like to know if there’s a way to “clear” the boosted words in a more direct fashion, or somehow turn off the boost “memory” between RecognizeRequests.
Thanks for any advice.