Confidence score uniformity

Is there a way to determine whether the training data has uniformly represented all the intents? I believe that this might be necessary to remove biases within confidence scores when similar words appear across multiple intents.

PS: This is my first question, so please let me know if I have missed any conventions of this community.