FAQ models improvement

I have some questions about FAQ models in Deeppavlov:

  1. Are you going to use Elmo or BERT in FAQ models?
  2. Are you going to create a FAQ model using info from users about wrong answers?


  1. Yes, we are currently doing some research on the use of BERT for FAQ models. Unfortunately, at the moment I can not announce when it will appear in the library. At the moment, our library has implemented a FAQ model using TF-IDF, perhaps this model is suitable for your task. http://docs.deeppavlov.ai/en/master/features/skills/faq.html
  2. At the moment there is no development of such a model.

Thank you, Pavel!
I already use this model.

Do you have recommendations about filling question-answer base for using with models listed at http://docs.deeppavlov.ai/en/master/features/skills/faq.html?