Attention in Conversational AI and Learning Reusable Patterns

by bendyBuson 12/4/18, 2:34 PMwith 2 comments
by ewooon 12/4/18, 9:36 PM

Great post guys! Sounds like a good approach. Looking forward to reading the paper to see how you modeled your training inputs (sequence of intents vs actions?). Also reminded me of an RL problem where you can consider your intents as states and the utterances as actions, where you can model a reward function based on the least amount of steps taken to reaching a fulfillment action.

by wrathagomon 12/4/18, 9:13 PM

Always impressed with the Rasa team's work!