Daffodil's parser could be encapsulated with the Apache Tika APIs allowing any DFDL-described format to be mined for text content in the Tika way.

Probably this would want to be schema-aware in that Tika events would not want to be reported for numeric content, but only text content.