http://tika.apache.org/ "detects and extracts metadata and structured text content from various documents using existing parser libraries". I use it all the time for input to solr.