Do you have any plan to realize mwer criterion?

[qzfnihao]

HI, as we know , mwer can reduce wer by 8% more on ctc and s2s. Does it have any play to realize it?

[tlikhomanenko]

We don't support this training right now, and we don't have this in our agenda for this half.

And pull requests are always welcome, so feel free to add this functionality!

[qzfnihao]

I have realized one mwer for ctc, just n-best mwer on cpu backend with beam search not prefix beam search. But when i finetuned the model trained with transformer+ctc, the wer in valid/test did not reduced. So maybe there are something wrong in my code that i did not be aware of.
I need some help to fix it. i want to know where can i get help for this? or paste the code here, in issue?

[tlikhomanenko]

you could post link to the code / code itself directly here, probably someone from us or from community will have a look at it and could help you.

[qgzang]

mark