Hi, Is there a way i can pull the training data of out_of_scope intent from txt file like a lookup. Data is huge to keep in intent. Rasa: 1.10
Any help is appreciable @Juste
Hi @shubh802. Can you explain in a bit more detail why it is problem for you to have an out_of_scope intent?
yes the problem is in out_of_scope intent the training data for it is in thousands is there a way we can pull training data for the out_of_scope from some lookup file like we do for entity extraction via lookup. Or any other way of handling the intent data.
@shubh802 Okay. I think I might know your answer to this, but why is it a problem that there is too much data under one intent?
@tyd We have implemented fallback action, so whatever intent does not gives data from the fallback action is covered by the out_of_scope. We have data around 5k which is coming from the out_of_scope. Hence my nlu file is becoming too big. And we are expecting more from that intent. Only one of my intent is too big for now.
@shubh802 It sounds like you are adding a lot more to the out of scope intent than you probably should. With task oriented assistants, out of scope includes everything that is outside of the domain of the task. Trying to add all of that into an out of scope intent is impossible. Instead, I would recommend focusing on adding out of scope examples that lie within the domain but that the assistant can’t handle right now. A great way to do this is to follow CDD and take real out of scope messages that users send your assistant as your training data
@tyd we are following the same process of CDD in training our chatbot. We just need to externalize our intent file for out_of_scope intent.
@shubh802 How many training data examples are in your other intents? 5,000 sounds like a very large number of examples for out_of_scope