When an intent is selected you can click on Disambiguate. This helps you find overlapping intents in your workspace. It will also expose the training examples that are ambiguous and that might need to be moved to other intents. This is similar to typical confusion matrices, except we'll expose a more focused confuse matrix to make the information easier to process.


Disambiguate is dependent on the active NLU engine you have active.


Disambiguating with KNN (k-nearest neighbor) does not require any training and helps you find intents that contain similar utterances as the one selected.

HumanFirst NLU or other NLU models#

Disambiguating with HumanFirst NLU or other NLU models requires a trained model and a recent k-fold cross-validation evaluation. The results will show utterances the NLU model had trouble classifying in the selected intent.