I am developing a chat bot for movie ticket booking.
if the user asks to "book tickets " it will show the languages of movies available. but when the user asking “book chicken” then also it is showing the languages. how to handle these type of out of scope questions. writing all these examples in out_of_scope intent is impossible.
so can someone tell me a solution for this.
Input → NLU processing pipeline → softmax → intents with confidences → pick up the higher one. This is the flow used in rasa. May be you could increase the minimum threshold and have some default fallback action asking user to be more specific. There may be other way outs too.
If i increase the threshold value the actions are falling into default fallback actions, if i use threshold as 0.3 the predicted actions have high confidence. but if i use threshold 0.4 its all going into default fallback. i am currently using my threshold 0.3 but even if i change the threshold to 0.4 , the out of scope questions have confidence greater than 0.5
I think your examples in nlu.yml training data are not very good, may be changing those could help in resolving the error
