I thought I would post a practical application, FWIW. I recently applied edit distance to a problem I was facing on my free dating site. Being free, we attract

I thought I would post a practical application, FWIW. I recently applied edit distance to a problem I was facing on my free dating site. Being free, we attract an inordinate number of scammers. I've been developing an arsenal of tools to help identify and track suspects in order to delete their accounts before they do too much damage. The most successful to date has used edit distance. Scammers tend to post identical or slightly altered messages to a large number of users. As you might expect, this has the effect of quickly polluting the valid ecosystem. The solution I employed was to delay sending messages between users long enough to capture spammer patterns and compare the messages to one another. I use the Levenshtein distance algorithm (which has O(n^2) performance). If I get a similarity over a certain threshold. I quarantine the message for further analysis. To work around the performance, I bound the problem by capping the number of messages used for analysis. It has been incredibly effective.