Sample documents and Examplotron

Whatever path you take, have a set of sample documents, and encourage others to create more.
Use cases at beginning Asking participants to bring sample documents forces them to look at the data closely and also lets you see the variations in participant expectations.
Sanity check throughout Sample documents offer a much more concrete view of the information than any schema specification ever will. They bring you closer to the data, show at least some of the variations possible, and let you test your schema.
Examplotron is a schema language that starts from the premise that every document is its own schema, and lets you annotate documents to create richer schemas. You can then convert Examplotron to RELAX NG and work with it as a schema.