How to create a test set

Hello everyone!

So, I am a “beginner” in Rasa, so this question might sound kinda boring… hope someone gets to read it!

I am trying to run the “rasa test” command. Before doing it, I used the “split” command to separate the data set. So, this is the result I can see on Jupyther lab:

Now, if I run the command: rasa test nlu -u --model models/nlu-20180323-145833.tar.gz the results I get are quite… optimistic. Every value (precision, F1 and so on) gets 1.00. In my opinion… something is wrong. Might it be that this command is not working properly?

@andreacirillo Welcome to the forum!

How much data do you have? Did you manually checked if the train and test data are different? If you have the same data in test as in train, it would explain the high values. However, that should not (normally) not happen if you used the split command correctly. If your data is very small and very simple, e.g. just a handful of intents, it might be that the model is really that good. Would be great if you could give some more information about the data you are using. Thanks.