Stories - The Training Data¶

Format¶

Here’s an example from the bAbi data:

## story_07715946                     <!-- name of the story - just for debugging -->
* _greet
   - action_ask_howcanhelp
* _inform[location=rome,price=cheap]
   - action_on_it                     <!-- user utterance, in format _intent[entities] -->
   - action_ask_cuisine
* _inform[cuisine=spanish]
   - action_ask_numpeople             <!-- action of the bot to execute -->
* _inform[people=six]
   - action_ack_dosearch

This is what we call a story. A story starts with a name preceded by two hashes ## story_03248462, this is arbitrary but can be used for debugging. The end of a story is denoted by a newline, and then a new story starts again with ##.

You can use > checkpoints to modularize and simplify your training data:

## first story
* _hello
   - action_ask_user_question
> check_asked_question

## user affirms question
> check_asked_question
* _affirm
  - action_handle_affirmation

## user denies question
> check_asked_question
* _deny
  - action_handle_denial

Visualization of Stories¶

Sometimes it is helpful to get an overview of the conversational paths that are described within a story file. To make debugging easier and to ease discussions about bot flows, you can visualize the content of a story file.

brew install graphviz
pip install pygraphviz --install-option="--include-path=/usr/include/graphviz" \
  --install-option="--library-path=/usr/lib/graphviz/"

As soon as this is installed you can visualize stories like this:

cd examples/concertbot/
python -m rasa_core.visualize -d concert_domain.yml -s data/stories.md -o graph.png

This will run through the stories of the concertbot example in data/stories.md and create a graph stored in the output image graph.png.

We can also run the visualisation directly from code. For this example, we can create a visualize.py in examples/concertbot with the following code:

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from __future__ import unicode_literals

from rasa_core.domain import TemplateDomain
from rasa_core.training_utils.dsl import StoryFileReader
from rasa_core.training_utils.visualization import visualize_stories

if __name__ == '__main__':
    domain = TemplateDomain.load("concert_domain.yml")
    stories = StoryFileReader.read_from_file("data/stories.md", domain)

    visualize_stories(stories, "graph.png")

Which will create the same image as the above python script call. The shown graph is still very simple, but the graphs can get quite complex.

If you want to replace the messages from the stories file, which usually look like _greet with real messages e.g. hello, you can pass in a Rasa NLU training data instance to replace them with messages from your training data.