OCR and pattern matching on text are computationally cheap and incredibly easy to do. For example, tax documents often bear the name of your government's tax authority, which presumably you are familiar with and can search for. They also tend to have years on them.