On the heels of yesterday’s image search post here’s another item connecting words and images, this one from researchers at MIT*. These guys have produced a “visualization of all the nouns in the English language arranged by semantic meaning**.” I had thought the English language looked like a large, disturbing bunny, but aparently it looks like an enormous mosaic of tiny colored blobs.
From their intro: “large-scale groupings correspond to broad categories such as plants or people.” Which lets us discover interesting trends, like that plants are green. That green blob at the bottom, floating around like Australia? Plants.
Then there’s this: “each tile [is] the average of 140 images. The average reveals the dominant visual characteristics of each word. For some, the average turns out to be a recognizable image; for others the average is a colored blob.” I clicked on dozens of tiles, and the average image was always a colored blob. This strikes me as analogous to taking all the synonyms for the word “person,” grinding them through an averaging algorithm, and claiming the average word for “person” is “aoviksv”. Which is to say, some things don’t make much sense, averaged.
So this is pretty useless, even by my low standards of what constitutes utility. What it really appears to be is an eye-candy outcropping of a larger, more meaningful research effort–machine recognition of objects in images. And who knows, maybe some fancy algorithm can make better sense of “aoviksv” than our tiny little brains.
Let me insert my standard caveat to digs at academia: what the hell do I know. These guys represent MIT. Errata represents… New Jersey. If that.