Wrong Prediction of Action

Hi Team

Embedding policy predicting wrong action every time

check story for different action

FIRST

  • match_won
    • utter_ask_game_type
  • inform{“game_type”: “cricket”}
    • slot{“game_type”: “cricket”}
    • utter_ask_game_name_cricket
  • inform{“game_name”: “ONE_DAY”}
    • slot{“game_name”: “ONE_DAY”}
    • utter_ask_date
  • inform{“date”: “last night”}
    • slot{“date”: “last night”}
    • action_match_won

SECOND

  • match_won
    • utter_ask_game_type
  • inform{“game_type”: “badminton”}
    • slot{“game_type”: “badminton”}
    • utter_ask_game_name_badminton
  • inform{“game_name”: “BWF”}
    • slot{“game_name”: “BWF”}
    • utter_ask_date
  • inform{“date”: “last night”}
    • slot{“date”: “last night”}
    • action_match_won

predicting the wrong action for story SECOND

actual prediction - utter_ask_game_name_badminton

executed prediction - utter_ask_game_name_cricket

Thanks

What is the slot type that you have used for “game_type”?

It’s categorical

In that case, may i see your “inform” intent and the structure of the slot that you have defined?

In the meantime I can also suggest creating a debug print in rasa with the following example:

In nlu.md

## intent:debug
- debug
- test

In stories.md

## debug path
* debug
  - utter_debug

In domain.yml

utter_debug:
  - text: '{game_type}'