Hello. I have multiple intents such as name, company name, school name. Now when I ask a question: What is your school name? If the user inputs a totally random new name of the school which is not present in my nlu training data for example St. Xavier. Now this can be classified anything out of name, company name or school name. How to handle these type of scenarios?
@mohit.bansal since these entities are all very similar to each other from a feature perspective, ideally you would just have one entity called “name” and fill a slot based on what the bot is currently asking for.
Thank you for the response @akelad
So, is it a limitation of RASA NLU that it can not handle intent classification properly when the entities are very similar to each other? Or is there any solution for this scenario?
Hi @mohit.bansal I think I misread your original post. You shouldn’t have separate intents for all these entities either, it should fall under a general intent “inform” ideally. This is a general limit with entity recognition (not rasa specific). The reason for this is that names generally don’t have a lot of well defined features. For example “Rasa” could be a company name as well as a persons name (in fact we know someone called Rasa 
).