I don’t know much about DIETClassifier, but i usually works with other machine learning algorithms and, when I try to evaluate questions that are used in the training corpus, they usually do not fail. However, it too common that the DIETClassifier fails when I introduce some exactly the same training questions. Is it happening also some of you? Is there any way to force RASA to be more accuracy with the training questions although it fails more with no previously seen questions?
Thanks, José Manuel