When I code in Haskell, the first and most important step I make is to create rich types for the domain I'm working in. If you disambiguate your types, your code becomes more readable and the compiler is a lot better able to help you out when you make a mistake.

For example, rather than having an Invoice with a status field of Text or String, I have a type InvoiceStatus = Paid | Unpaid | Void. From there, you can write a very precise parser that takes exactly what you want and nothing else. When I apply this approach to each component of the thing I need to parse (or do it generically even), I don't really find myself having to munge javascript types into my data structures.