Reputation: 1572
Doubts:
It seems that interactive learning completely ignored pre-defined stories.
How do I create multiple stories with interactive learning? Would have to make a way "happy", generate the file, start a new training and then join the files? I tried to fix multiple paths, in the file generated by the end it seems to have become a single story.
I'm working with Rasa Core, to create context conversations in a chatbot. First I did this basic tutorial and then tried to make a story file using markers, to make sure I understood how to use markers.
stories.md - It looks like this:
## Story Hi
* greet
- utter_greet
## Story Good Morning
* mood_great
- utter_happy
## Story bad_day
* mood_unhappy
- utter_cheer_up
- utter_did_that_help
> marking_bad_day
## Story bad_day Best
> marking_bad_day
* mood_affirm
- utter_happy
## Story bad_day nothing changed
> marking_bad_day
* mood_deny
- utter_goodbye
## say goodbye
* goodbye
- utter_goodbye
But by testing the conversations with the robot, he does not seem to follow the markers. So I mixed the learning of the third tutorial, which is using interactive learning. This seems to have solved my problem, I really managed to make the bot follow the flow wanted, unfortunately with a great effort of learning.
New stories.md:
## Generated Story 5325363759313905425
* greet
- utter_greet
* mood_unhappy
- utter_cheer_up
- utter_did_that_help
* mood_affirm
- utter_happy
* greet
- utter_greet
* mood_unhappy
- utter_cheer_up
- utter_did_that_help
* mood_affirm
- utter_happy
* greet
- utter_greet
* mood_unhappy
- utter_cheer_up
- utter_did_that_help
* mood_deny
- export
Upvotes: 1
Views: 593
Reputation: 1285
after you finished online training, type 0
to export dialogue and then enter filename (eg. story01.md
). Now open trained story (story01.md
) and copy/paste to data/stories.md
or wherever you stored your stories.md
training file.
Can you provide your train_online.py
file maybe?
Upvotes: 2