Using rasa 1.1.4 with supervised_embeddings I create a project with rasa init --no-prompt I run through the project with rasa shell, once I go through the main story it does not repeat and the bot defaults to action_listen although correctly recognizing ‘hi’ as intent greeting. How do I set the bot so after finishing any story it is able to rerun on jump into any other story? Because this is the default behaviour of a default project is there something that I am misunderstanding about how stories work?
Hi Martin,
the bot is supposed to give answers that make sense in the context of the conversation based on the stories it was trained on. The example bot does not have any stories in which intent greet comes after a path was concluded. In general the answer is therefore to add more training data and add more variations that cover the cases your both should know (check out Checkpoints).
Thank you @imLew that makes sense, it just felt like a strange default behaviour. I understand the concept of stories for managing nested intents but I struggle to see their benefit in cases where a story could be just broken down into separable pieces of greeting story, goodbye story, i am bad story, i am good story, which in my mind would yield a more robust behaviour of the chatbot.
While in most cases I wouldn’t recommend it, you could probably achieve such behaviour by decreasing the max_history parameter in the config (Policies).