The purpose is to check whether one or more of your text files in your corpus doesn't belong. This could be because
•it has got corrupted so what used to be good text is now just random characters or has got cut much shorter because of disk problems
•it isn't even in the same language as the rest of the corpus
•it contains too much boilerplate text
•it is relevant to a particular area of enquiry
The tool works in any language. It checks corruption by using a known sample of good text (in whatever language) and comparing that good text with all your corpus.
See also : How to check corruption