I’m using Rasa NLU and Rasa Shell.
I’m having trouble when the user sends the same question as the one that we trained the bot with, but without a space between the final word and the question mark. The Rasa doesn’t recognize the intent anymore.
This shows that your model is fit on the pattern [space]?
Try to vary your examples. when you fit a model, it is basically looking for patterns in a sentence, here it found that almost all(infact) all your examples have the same pattern and this is taken up as an input feature for your classifier. Remember that the pipeline tokenises every word in your sentences and the tokeniser most likely is based on whitespace. meaning the whitespace between the last word and the question mark is considered another token and it becomes part of your input feature set.
Vary your examples to create more co-relation between the intent and the examples