Hi., I am building a legal assistant chat bot. Let’s assume in a situation where client is clearing her mundane doubts with questions like ‘what is a bail’, ‘how much hiring a lawyer cost?’ or ‘how to file a divorce’ then these queries do not fall under any predefined intents but nlu_fallback.
My question is how to handle the repetitive sequence of [ nlu_fallback → action_handle_fallback ]+ (+ indicates 1 or more occurrence in Regex ) scenarios in story.yml ?
Hi, I’m not entirely sure what you mean by
repetitive sequence of [ nlu_fallback → action_handle_fallback ]+ (+ indicates 1 or more occurrence in Regex ) scenarios
If the goal is to just tell the user “We’re not able to handle that question”, then define an out_of_scope intent with all the examples you gave, and add a regular rule or story:
rules:
- rule: out-of-scope
steps:
- intent: out_of_scope
- action: utter_out_of_scope
(See: https://rasa.com/docs/rasa/fallback-handoff#handling-out-of-scope-messages)
If the goal is to handle all these mundane messages with a single intent: how about defining a single intent mundane_questions which groups all of these mundane questions?
Then you could then handle the regex in a couple of ways:
# With entities
# One story per entity
- story: Mundane question 1
steps:
- intent: mundane_questions
entities:
- some_entity_1: some_value_1
- action: utter_answer_mundane_question_1
- story: Mundane question 2
steps:
- intent: mundane_questions
entities:
- some_entity_2: some_value_2
- action: utter_answer_mundane_question_2
# Etc
or with an action
- story: All mundane questions
steps:
- intent: mundane_questions
- action: handle_all_mundane_questions_with_code_which_deals_with_regex
That would leave you the possibility to use nlu_fallback for something else.
If you really want to use the fallback mechanism, it’s possible to do something similar to the out_of_scope suggestion, again by writing a custom action (see: https://rasa.com/docs/rasa/fallback-handoff#nlu-fallback)
Hope that helps