Hi Rasa community,
I try to fill two slots in a form from_text but in rasa interactive it always misclassifies the intent (actually it should not classify any intent as the message is only filled from the user response) and the slots keep staying empty. I have found this issue already on GitHub (https://github.com/RasaHQ/rasa/issues/10053) but there does not seem to be any solution/explanation for it.
This is how my code looks like:
slots:
nick_name:
type: text
influence_conversation: true
mappings:
- type: from_text
conditions:
- active_loop: start_form
requested_slot: nick_name
favorite_dish:
type: text
influence_conversation: true
mappings:
- type: from_text
conditions:
- active_loop: start_form
requested_slot: favorite_dish
forms:
start_form:
required_slots:
- nick_name
- favorite_dish
This is the corresponding story:
stories:
- story: happy path cold shower start
steps:
- intent: greet
- action: utter_greet
- action: action_start_conversation
- intent: affirm
- action: start_form
- active_loop: start_form
- slot_was_set:
- requested_slot: nick_name
- slot_was_set:
- nick_name: hansi
- slot_was_set:
- requested_slot: favorite_dish
- slot_was_set:
- favorite_dish: rice
- slot_was_set:
- requested_slot: null
- active_loop: null
- action: utter_same_dish
- action: action_start_topic
I do not need to get the actual entities or intents. The user can say anything for their nick_name and the bot should just record it. Same for fav_dish. The bot will just always answer “good choice”. It is to test the from_text feature but the slots do not get filled. I hope someone can point me to my mistake… Maybe I do need to classify an intent? But how would I then provide training data for the intent as I do not want to give examples for the nick_name as it could be literally anything from “Banana” to “John” to “Superhero”?
Many thanks!