Once you have trained the model , you would want to improvise it this will often be done by adding more training data.To do this,you need to identify the type of conversations that your model can handleand those that require more training data.
Interactive learning allows you to provide feedback on errors.. Suppose you are building a restaurant search bot and the bot asks “How may I help you?”. The user replies “Please find Chinese restaurants in Mumbai”. In this case, the slots of both location and cuisine are filled. However, the bot may again ask ‘in what location’ or ‘what’s your cuisine preference’ mistakenly. You can correct this here and guide the bot to instead ‘search for restaurants’ as the next step.
We recommend that you create stories using interactive learning, because if you type these manually, you would probably forget to add all of the slots
You can read more on interactive learning from Rasa’s documentation.