doykle

Apache Tika is a text extraction toolkit. They store a wide selection of file types for their parser tests: https://github.com/apache/tika/tree/master/tika-parsers/src/...

For large sets of some common media types take a look at the govdocs1 corpus: http://digitalcorpora.org/corp/files/govdocs1/by_type/

For an odd example, sometimes a google search will turn something up.: https://www.google.com/?q=filetype:xlsx