only use MultiProjectImporter?is there another way to do multi law court?In my rasa, there some nlu and response is common parts, and there are also different nlu and response for different courts.
This is the core config:
importers:
I want that each folder is different court, and the nlu.yml of different court belong to respective court.but after I get the train model, it seems like that the rasa can response to each court.When court is lawbot_shushan, I want it ignore the content of other folder. how to do it? how to ignore else?
I think you have a couple of approaches you can take. Either (1) generate a single modle and use a slot to specify the context or court that the conversation or intent is about; or (2) create separate models for each court (you’d have separate runtime bots for each court).
It sounds like you are taking the 2nd approach. In that case, I’ve seen two approaches for generating multiple models from a single repo. One is to do this with scripts (and part of a CI/CD pipeline if you want) that creates a data directory from the combination of the common and court specific training data for each bot and trains them separately. The other approach I’ve seen is to separate git submodules for each court (still have separate build scripts).
Use channel in Rasa? there 600 courts, some intents repeat in other court, but I can’t delete it,beacause 600 courts takes long time. so when I stand in one court, it recognize the intent is the courts,not the intent of other court, such as court A : intent :goto movie 1 ,text: see a movie; court B : intent goto movie 2 ,text: see a movie; court C : intent goto movie 3,text: see a movie; when I in court C ,it can recongize the intent is intent goto movie 3, not the similar intent intent goto movie 1 and intent goto movie 2
I use the tracker.get_last_event_for(“user”, [], 2) ; tracker.current_state() ; tracker.events . but it only display 10 similar intents ,how to get all similar intents. or the numbers of intents is 600,is same with numbers of courts.