- added a unicode ASCII-fiying addition to the fingerprinting functions
- removed all distance functions for kNN that didn't seem to do anything useful
- added the ability to indicate what value to use as cluster centroid by simply clicking on it
(this is useful for those names that have non-ASCII chars that might not even be on your keyboard.. and cut/paste is error prone/cumbersome)
- added a 10x multiplier to the PPM compression distance which makes it more aligned with the levenshtein ones
- made sure that we construct a phonetic fingerprint for the whole string and not just the beginning subset
(performance is still not ideal but it's now reasonable)
git-svn-id: http://google-refine.googlecode.com/svn/trunk@268 7d457c2a-affb-35e4-300a-418c747d4874
- added a few of the more powerful distances
- fixed a bug in the VPTree builder (although is still not working as I expect it to)
git-svn-id: http://google-refine.googlecode.com/svn/trunk@248 7d457c2a-affb-35e4-300a-418c747d4874
- added the secondstring libraries that contains all sorts of useful string distance functions
- added a java arithmetic coding library (used to implement a string distance based on PPM arithmetic coding)
- added the vicino kNN string clustering library (from MIT's SIMILE)
git-svn-id: http://google-refine.googlecode.com/svn/trunk@181 7d457c2a-affb-35e4-300a-418c747d4874