Hi
I am currently working on a RASA NLU pipeline that predicts 2 intents and 20 entities. Is it possible for me to calculate Logloss and AUC at overall model level?
there is confidence score generated for every entity and intent predicted but I am not sure if that is the probability of prediction as given in this link: http://35.196.60.7/docs/nlu/0.13.3/fallback/
Does it even make sense to analyze Logloss and AUC in NER in RASA and how to do it?