- Can I implement it so that the payload of each button has the /intent_value as its payload? Accordingly, I assume that the NLU would then skip the evaluation and simply predict the intent, is that correct?
Yes, you can mention the intent, for example:
responses:
utter_greet:
- text: "Hey! How are you?"
buttons:
- title: "great"
payload: "/mood_great"
- title: "super sad"
payload: "/mood_sad"
And you need to write the story for the same.
- If the answer format for a question is a button, is training still needed for the chatbot to know which intent is meant? Or can I also use this to reduce the training and thus the size of the model?
Yes, you need to train the bot, as intent will be trigger based on story and conversation flow.
- If only buttons are allowed in the whole conversation “menu bot”, would RASA be a good choice at all or should I use a framework that doesn’t use training at all to keep it slim?
Well, that depends on your requirement and use-case. Normally, we consider both Text and Button approach, other wise user is bound to answer based on buttons.
- Is it somehow possible to disable the free response format, i.e. the text field where responses are entered, so that the user is “forced” to respond with a button?
Yes, you can do it but its much depends on your front end.
I hope this will help you.