Analyzing intent error report from rasa test nlu (Ambiguity in NLU Training Data)

Hi, I am creating an intent classification model in Rasa. And as I test the model with “rasa test nlu”, it gives me “intent report.json” in result. After analyzing the report I find ambiguities in my NLU training data and I resolve all of those in my NLU file. I retrain my nlu and repeat the process again. And each time, the result shows new ambiguities in my NLU data.

Is iterating over this process a good idea to better my training data?

Hi @NIlimaN,

How did you resolve this ambiguities in your NLU file, is it manual or using some tool ?

As discussed in previous sections, NLU’s first task is intent classifications. So, you need to tell your Chatbot what are the intents to expect.