I wonder what you folks from rasa think about training data generators like Chatito. I know it doesn’t replace real conversations, but do you think they can improve NLU performance? Or do the ML models learn these abstractions anyway?
Welcome to community!
Yes, they improve your NLU performance, especially at the beginning when building the MVP.
At RASA channel on YouTube, you’ll find a series called Masters class, there they even suggest the use of this type of resource to improve your bot.
I did read the Masterclass, but did not encounter that part. Could you be more precise about where to find that passage on website or in the videos?
man, I let the master class playing here in background but I wasn’t able to pinpoint the part.
But I swear that they mention there (or at least mentioned), because I got to know Chatito from there.
I’ll try to watch latter with more attention.