Creating NLU data for names that should not be recognized

I want to make a chatbot where the user tell their name, but it is not important for the bot the recognize it. This could then be any name. I am wondering what the best practice is then: Should I just keep the names out of the NLU data? or use a sinlge name? or is there a way to do set up a genereal entity for this?

hi @heineaabo - if I understood you correctly, you have in mind the use case that a user says “hi, I’m Heine” to the bot, and that is classified as intent greet and the name is NOT extracted as an entity?

You can certainly do this with training data that does not include the name as an entity [], e.g.:

- hello
- hello my name is Heine

and just make sure you don’t have an entity recognition component in your pipeline that’s pre-trained for names. The SpaCy entity recognizer is the only standard Rasa component which does this, so just avoid that one.

Yes that was my intention, but I also want to prepare the bot for all types of unknown names so:


  • hello
  • hello my name is Heine
  • hello my name is Alan …

My question is then if I need to type an example for a broad range of names or if there is a more efficient way to do so?

I also have a greet intent where the user do not say their name, and I want to distinguish these two intents.