KewMann interview question

How do you think external unstructured data can be used to augment internal data?