Internal NLU - testing utterances

I really like how the Dialog Engine is designed! I do struggle though with the builtin NLU.

I find it difficult to understand why some utterances aren’t identified. This seems to be the case when there are two high-scoring intents and it might be ambiguous.

It also incorrectly classifies occasionally if there’s a specific word in an intent that has an exact match on only that word in another intent.

I can’t find how to enable Debug - or where to see the information. I have enabled NLU in the global Debug tab in the UI. The songs described here (https://botpress.com/docs/main/nlu) don’t seem to exist (or I can’t find it) in version 12.5.

I also can’t find where I can set a confidence threshold. Any pointers?

How do I test how utterances are classified? Or is the internal NLU not recommended yet?

1 Like