onPartialResult callback and grammar

[matanel-6over6]

First of all it works great and fast.

1. onPartialResult callback called several times if I say one word, why?
2. I have grammar(json array) and If I say a word that is not on the list, he still returns one of the words on the list. There is no option to identify that there is no match at all? Thanks for the help

[nshmyrev]

onPartialResult callback called several times if I say one word, why?

Partial results returned 5 times a second to provide you real-time updates.

I have grammar(json array) and If I say a word that is not on the list, he still returns one of the words on the list. There is no option to identify that there is no match at all?

Add a special [unk] word to the list

[matanel-6over6]

Thanks for the fast answer. This is my json array: ["yes","no","none","next","zero","one","two","three","four","red","blue","green","yellow","white"]

I say 'five' and i get four

[nshmyrev]

You need to add "[unk]" to the list

[matanel-6over6]

["yes","no","none","next","zero","one","two","three","four","red","blue","green","yellow","white", "[unk]"] like this? I say five and from onPartialResult i get four but from onFinalResult i get [unk]. why?

[matanel-6over6]

I sat seven and i get in the first onPartialResult "zero" and after that i get [unk]

[nshmyrev]

Yes, partial results are preliminary and not very reliable

[matanel-6over6]

Is it just the first time? I can count on it the second time and onwards will be [unk]?

[matanel-6over6]

There is an option to get reliability or score in onPartialResult?

[nshmyrev]

You can find more accurate model, yes! In general it is very hard to spot words in realtime, you'd better have some delay.

[matanel-6over6]

Thanks a lot

[matanel-6over6]

Last question, where can I find a more accurate and slimmer model?