I use a small grammar, and in order to reject OOV I use a rule similar to the <garbage_loop> rule found here. But in my rule, I don’t add the plus sign at the end. So my rule detects a single phone, not a sequence of phones. And It gives good rejection.
When I use a phone loop (i.e., by adding the plus sign), I get very high WER, and the rejection is not better. So I wonder what are the uses of a phone loop.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I use a small grammar, and in order to reject OOV I use a rule similar to the <garbage_loop> rule found here. But in my rule, I don’t add the plus sign at the end. So my rule detects a single phone, not a sequence of phones. And It gives good rejection.
When I use a phone loop (i.e., by adding the plus sign), I get very high WER, and the rejection is not better. So I wonder what are the uses of a phone loop.
Phone loop helps with longer phrases, maybe your input is usually short. It also helps to add top 100 frequent words of the language.