Hello, I see some video on youtube and found there are a way to train the chatbot is ‘rasa train’ and the training begin with the data in .yml files.
But i was not sure about if it’s possible to provide chatbot to know that the predicted message is incorrect at their current conversation [in test phase] . and Can the chatbot know they predict wrong sentences? and Can they train the model that act correct to the customer chat themself without ‘rasa train’?
[think about reward system in reinforcement learning. Reward is given to agent if the action is good: +1,-1. And How can they get reward -1?]