On Tue, Mar 20, 2012 at 23:09, Ken Takusagawa II ken.takusagawa.2@gmail.com wrote:
One more note: the Soundex and Double Metaphone algorithms may be useful for determining if two words sound alike.
True. But we don't care nearly as much about homophony as about synonymy.
Homophony is a heightened concern for things primarily intended to be conveyed over the phone (eg PGP fingerprints). Our primary concern is rather that the things can be memorized — and memory, at least for phrases rather than isolated words, is semantic.
E.g. we would not drop 'goat' just because we have 'coat', even though the two are very similar phonetically (just one voicing difference).
While we're at it, homography is also a lesser concern — it'll mainly come up when ensuring that parsing is unambiguous. (Granted, that's a significant caveat. :-P)
- Sai