Hi All,
Just reading this blog post from @amn41 . It’s really interesting.
Is there any progress on this one? Do we really need to move away from intent?
Cheers
Hi All,
Just reading this blog post from @amn41 . It’s really interesting.
Is there any progress on this one? Do we really need to move away from intent?
Cheers
thanks @welly87! Yes, we’ve been making lots of progress on this but still some way to go. In Rasa Open Source 2.0, intents will become optional. Check this post What’s Ahead in Rasa Open Source 2.0
Wow, that’s mind blowing…
so it will become end to end training as a whole? how about explainability of the whole pipeline?
because if we separate intent classification and dialog policy, would be easier to debug and reason?
is there any way to overcome this limitation?
cheers
definitely not! Rachael’s points are all super valid. We’re still working out the details, but the idea is that in Rasa 2.0, you still can use intents, but you don’t have to. You can also mix and match, so you can have training stories where some messages are associated with an intent, and some don’t have an intent at all.
ah right… so that’s what you mean that intent become optional… thanks.