While we are learning about cognitive services and playing with the CLU feature, it would be useful to be able to change the success threshold of CLU when searching our intents.
We are finding that the current threshold picks up unrelated smaller words and nonsense strings which can be confusing when configuring and testing.
The vast majority of these false positives sat in the 70-79% range so I would be interested to play with the threshold to test our flows with a higher threshold as I think with some datasets the number may need to be different to the default. From my testing, 80% would be great for the flows I'm working on but having it configurable allows us all some flexibility.
Generally, when it comes to whether a project flow fires from CLU, it's may be worth us erring on the side of caution - i.e. force the user to do more work (via utterances) for Tenjin to hit (a potentially higher) threshold rather than allowing it to fire more often than it should via unrelated intents.
