Many similar intents in multiple scenes

Hello everyone, because there are many scenes, there are many different keywords and entities in the sentence between scenes. The text of each sentence is also short, which leads to confusion in the intent of identification in a model. How can this problem be solved? Thanks in advance.(similar to Contextual Intent Classification · Issue #1388 · RasaHQ/rasa_nlu · GitHub

I’m not sure what you mean, could you provide an example?

For example, for ‘okay’, it can mean ‘certain’ in a different dialogue in a scene, and it can also mean ‘goodbye’ in another scene. My current plan is to merge similar intents, but it causes the next trouble. Because I merged a lot of intents, I put the difficulty of the dialogue on the core (core needs to respond differently to different entities according to the same intent), I improved a lot of train policy, and even deliberately let it fit. (Because the story itself is not predicted correctly), but the core response after the intent is still wrong, and does not follow the content of my story. I guess the value of slots is only 0 and 1(I mean - slot{“account_type” : “premium”} and - slot{“account_type” : “basic”} have the same featurizers), so it seems to have nothing to do with different value of the slots . how to print the featurizers of slots value with Interactive Learning.thank you for reply.

You just have to write different stories for this in different contexts. As for slot featurization, there’s different types of slots which behave differently. Please take a look at our docs: Slot Types

Thank you.

1 Like