How Do You Handle Ambiguity in Natural Language Processing?

Natural languages are full of ambiguities and nuances that can challenge even the most sophisticated NLP models. What techniques or models have you used to deal with ambiguity in your NLP projects?