What methods do you use to clean and prepare unstructured or messy data before performing an analysis?