This article is interesting:
Combining methods to better search for words that sound like other words when searching a MySQL database
This kind of idea can be quite handy when combining feeds of data from different sources.
For example feed x may tell you the specifications of a product, feed y may tell you the the purchase information for a product and each feed may come from a different source e.g. feed y may come from a manufacturer and feed y may come from an affiliate network.
If the manufacture has a slightly different name for the product to the affiliate network (in this example), and the name is the only possible field on which to combine the data sources then it may be difficult to associate the two entries.
Methods such as the one illustrated allow us to make an varying levels of intelligent guess of what matches up with what, and as such can massively assist what could otherwise end up being an entirely manual matching process.