How do we store hyphen containing words in slots?

I have an input policyNo. In our production policy numbers usually contain hyphen. As of now RASA NLU is not able to extract the word containing hyphen. Please suggest. I am using spacy en pipeline.

what does your training data look like? it’s entirely possible that the spacy tokenizer splits on hyphens

you could use the whitespace tokenizer instead, which is part of the supervised_embeddings pipeline