Hello! We’re developing a complex scenario using complex forms and thought of seeking some help in the “formbot” example, it seems particularly fitting. But despite having run each step (including actions and the duckling server via docker, everything seems operational), the bot’s intent recognition seems to fail miserably, even fails to capture the intents that are defined as the very training data!
I’m using Rasa 3.1.0 installed from rasa[spacy] via pip (Ubuntu, Python 3.9.5, everything inside a virtualenv).
Bot loaded. Type a message and press enter (use '/stop' to exit):
Your input -> hello
Sorry, I didn't understand you, please try input something else
Your input -> hi
Hello! I am restaurant search assistant! How can I help?
Your input -> i need a restaurant
Hello! I am restaurant search assistant! How can I help?
Your input -> looking for a restaurant
Sorry, I didn't understand you, please try input something else
Your input -> im looking for a restaurant
You are welcome :)
Your input -> im looking for a restaurant
You are welcome :)
A look at the interactive mode to see that looking for a restaurant is captured as the thankyou intent:
Type a message and press enter (press 'Ctrl-c' to exit).
? Your input -> im looking for a restaurant
? Your NLU model classified 'im looking for a restaurant' with intent 'thankyou' and there are no entities, is this correct? (Y/n)
I’d be grateful if you could help find the cause if this odd behaviour… perhaps some misconfiguration or this is an experimental version…