Did you ever want to see a graphical representation of the semantic relationship of all the nouns in the English language? Yup, there it is. There are 79 million or so images synthesized down and mapped to 53,463 nouns. It’s part of a project to train computers to recognize objects and scenes in images. The theory is, apparently, that if you have a hundred pictures of tables, and a hundred pictures of chairs, and you form a composite of the pictures by averaging them together, you’ve created a unique index of "table" and another unique index of "chair". I have to suppose it will work in a general way, but if you click around the dataset there, you’ll see some interesting things (no pun intended). For entry #28543, which is the word "hatemonger", it looks one of the pictures of the subset is that of Michael Moore, and another is of Lou Dobbs. Even without being told, one could guess that the image database was drawn from the internet. I was somewhat surprised when I clicked on the extreme right side of the next to bottom row, which turned out to be entry #53134 and the word "bang". The banging that’s going on in those pictures is not what I would call a noun.
via: Andrew Sullivan
| M | T | W | T | F | S | S |
|---|---|---|---|---|---|---|
| 1 | ||||||
| 2 | 3 | 4 | 5 | 6 | 7 | 8 |
| 9 | 10 | 11 | 12 | 13 | 14 | 15 |
| 16 | 17 | 18 | 19 | 20 | 21 | 22 |
| 23 | 24 | 25 | 26 | 27 | 28 | 29 |
| 30 | ||||||