Details about conversational-ai-workshop-18


What is the reason to set the slots as unfeaturized ? (conversational-ai-workshop-18/domain.yml at master · RasaHQ/conversational-ai-workshop-18 · GitHub).

If set in such a way, how the embedded policy know what has already been set so far (unfeaturized -> no memory) ?


it knows it from its lstm c memory and attention

Thank you for the answer …

What would be the impact on the training/accuracy if the slots were set to “text” ?

it is more complicated to generalize for chitchat and did_that_work at any place