Suppose I have topic A,B,C,D,E,F,G,H and any combination is possible is there a better way of having multiple intents classified, as the training data will be prohibitive for all these combinations.
Hi Dip, do the corresponding stories differ for every intent combination?
For example, if A+D and B+D follow the same story in Core as intent D, it would make sense to add examples of A+D and B+D directly into intent D instead of using multi-intents.