While the discussion has been about comparing this technique to words and text, it is also a way to handle leading wildcard searches on part numbers. I have done something like this by building an index of all reasonable suffix substrings of a part number, clipping off one leading character at a time until there's nothing "significant" left. Then if a user wants to search for, say, '%74%138%', I clip off the leading wild characters and do a LIKE on the suffix index. (I could further optimize by guessing that '138%' would be a shorter scan than '74%', and applying the original '%74%138%' pattern to the original part number.) I quickly get all the matches like 'MC74AC138N', 'SN74LS138D', etc. I don't imagine that FTS would be applicable to this problem.

While it sounds like I'd be multiplying the size of my index by the number of characters in a part number, it's really not so bad. Short part numbers have fewer suffix substrings, and you can add some restrictions if you disallow putting a wildcard character in the middle of certain sequences, like a series of digits. For about 125,000 part numbers I have less than 600,000 suffixes to part number mappings.