#462: Soundsearch

Search has become fascinating for two reasons: the recent availability of an abundance of interesting stuff (if only you knew where it was) and the possibility of automating its retrieval.

One of the biggest problems, it seems to me, is that we currently aren’t really making use of all the tools available to us. It’s still hard work to find things visually or kinaesthetically in a reliable way, but sound offers a more tractable approach.


Today’s invention is to store a phonetic representation of every word in a document or webpage. This would enable a search by sound to occur in which we could say “find me things that sound like fandango” or even, in wildcard mode “…things that sound like blahblah_blah_blah_ford.” This is pretty much how I try to retrieve people’s names from my own memory, and it seems particularly good for things whose names don’t mean anything but whose phonemes may be identifiable as words …eg brands like coca-cola or phentermine or whatever.

This makes me wonder if, by populating one’s website with spam brandnames, you would be less bothered by the makers of smarter, targeted spam (the kind of message about which you think ‘is that spam or is it a genuine message?’ ). These people know it’s a bad idea to pepper spam blockers with data that helps them extract a characteristic pattern and thus defend themselves.