Hi All,
I am facing a weird issue using rasa x interactive learning. after collect some conversations copy to rasa x interactive learning I noticed I start to get failing stories after run rasa test.
giving a deep look up to the stories I found the one generated by rasa x are different from the ones generated by command line interactive learning in how the FormActions are writing.
If I am using pure cmd IL (Interactive Learning) no failing stories, the same if I’m using pure rasa x IL So my conclusion is both are ok alone, but if you mixed both type of stories you start to get prediction issues in the stories.
Is there any way to prevent this situation? any way to convert those stories to be compatible between them, it takes too much time to manually go one story at a time.
Example with rasa x Interactive learning:
* agendar_servicio{"location":"miami","time":"2019-06-14T10:30:00.000-04:00"}
- slot{"location":"miami"}
- slot{"time":"2019-06-14T10:30:00.000-04:00"}
- agenda_form
- slot{"location":"Miami"}
- slot{"time":"Fri, 14 Jun 2019 14:30:00 GMT"}
- slot{"requested_slot":null}
- slot{"requested_slot":null}
- action_next
Example with command line Interactive learning:
* agendar_servicio{"location": "miami", "time": "2019-06-14T11:30:00.000-04:00"}
- slot{"location": "miami"}
- slot{"time": "2019-06-14T11:30:00.000-04:00"}
- agenda_form
- form{"name": "agenda_form"}
- slot{"location": "Miami"}
- slot{"time": "Fri, 14 Jun 2019 15:30:00 GMT"}
- form{"name": null}
- slot{"requested_slot": null}
- action_next