Training vs Testing about chatito HOT 4 CLOSED

Hi @harirehm ,

Generating testing dataset is optional, so you can just pass the training entity argument and it will be ok if you just want to train your rasa nlu model. The problem with only having training data, is that you will have to evaluate the model manually. To make things easier for evaluation, you can generate a testing dataset too.

You can use the testing data to evaluate how accurate the model is with data it never trained with, this way you can compare the accuracy of your models, makes them easier to maintain and improve. As mentioned, is not required, but is important to be aware of the accuracy of a model so you can get to the highest accuracy. Sometimes just adding more and more data, can degrade performance of a model because of overfitting, this kind of issues can only be detected if you have a testing dataset.

from chatito.

will keep this issue open to remember to improve docs on this topic

from chatito.

Clean!

from chatito.

Done. Updated documentation about testing dataset. Thanks

from chatito.