Hi,
I wonder whether there is a chance to use some question-answering language model for general conversation (i.e deepset/roberta-base-squad2 路 Hugging Face) instead of feeding the conversation into NLU and domain (this will take forever to write them one by one 
) and not only use the transformer for classification (i am not into giving the class), but also for response generation since what I want to build is not just a customer service chatbot or something similar but a personal assistant.