As of this writing, Google Sets (http://labs.google.com/sets) is one of the cooler things in the Box Of Cool Things that is Google Labs (labs.google.com). Google Labs is a collection of web projects created by Google staff members as a way to test out new ideas, try out new technologies, and let off some of the incredible amount of inventive steam that builds up when you have so many bright Ph.D.s employed by your company. Google Labs has lots of neat stuff, but for my (admittedly puny) money, this leads the pack.

This is how you use it:

You start out with a classic Google plain white page with five one-line text boxes and "Large Set" and "Small Set" buttons. You fill in a few members of any set of terms into the boxes and then click one of the two buttons, and doodly-doodly-doop!, Google will then try to add new items to the set to match what you entered. For example, if you enter a few colors, you'll more than likely get several more colors when you press one of the buttons. This works both for simple kinds of things, and some things you wouldn't expect. Some things it doesn't work on. Part of the fun is finding out which works and which doesn't. You don't have to specify the category you had in mind, though sometimes Google Sets might pick one you hadn't figured on, or even mix two closely-allied sets together. Still, just the thing for Bluetooth-enabled SAT takers.

Google seems to use its massive (yet not yet omniscient, despite its cataloguing of Everything) web page database to figure out which terms match with others. It does a fairly good job of weeding out irrelevant things, considering the difficulty of the task. It does make mistakes sometimes, though, especially with the large sets. It's also possible for a set to come up with nothing but the things you started with. If you hit the "Small Set" button, which typically gives less than 15 items, you'll be given the option to "Grow" the set afterwards into a Large Set, and vice versa. You aren't guaranteed to get any items beyond what you entered with even a Large Set, let alone a small one. The set you get by growing a Small Set seems to be exactly the same as the one you would have gotten if you had chosen the "Large Set" at first. It seems that, once in a while, it is possible for a Large Set to be missing elements from a Small Set produced by the same list. The order of the elements in your results may not be significant, but you usually get the items you entered before any others.

Final notes before we get down to bidness. Set items can contain more than one word, and when the results come up each term will be hyperlinked to a Google Search for that term, which is even more nifty.

Some experiments follow, and these are their circumstances:
- Mozilla 1.4 RC1
- Windows XP SP1
- Sony VAIO PCV-RX550
- Double Caffeine

Seed: Charlie Brown, Snoopy, Linus
Small Set Adds: Lucy, Schroeder, Peppermint Patty, and Sally (hey not bad!).
Large Set Adds: Woodstock, Joe Cool (oops, that's a Snoopy fantasy and not a character), Pig Pen, Franklin (didn't he go the way of Eudora?) and Marcie.
Comment: It's summer already, but where are the Re-Runs?

Seed: Bush, Clinton, Bush (heh heh, this'll trip it up), Reagan, Carter
Small Set Adds: Nixon, Ford, Eisenhower, Kennedy, Johnson, Truman. (It got rid of my second Bush.)
Large Set Adds: Harding, Jackson, JFK (Jiffkuh?), Bush Sr (hey impressive!), Buchanan (what th-! Oh yeah, James), Kissinger (oh my god!), Lincoln, McGovern, Agnew (alas poor Spiro), Mitchell(?), Ehrlichman(??), Nixon Pat (um okay), Goldwater, Halderman, Zhou Enlai (?!), Hoover, Habib (!), Hiss, Sato Eisaku, Dean (Is this a James?), Taylor, Polk, Cleveland, F Roosevelt, Pierce, Jefferson, L Johnson, B Harrison, Tyler, Madison, A Johnson, T Roosevelt, Arthur, Roosevelt and Hayes.
Comment: This one didn't turn out as well as the Peanuts characters, but at least most of them were close. Something weird happened here: I accidentally mistyped one of the Bushes as "Bust" at first (I dunno, I guess it seemed fitting), but Google removed the word from the results. Pretty slick. On the other hand, I got different (and actually slightly better) results that run, including Taft and Washington. Removing one of the Bushes, too, gave different results. The set pixies appear to be mischievous. All results seem to be missing any periods, which is probably some arcane Perl summin'.

Seed: Pepperoni, Hamburger, Sausage
Small Set Adds: Cheese, Broccoli, Spinach, Onion, Shrimp, Pepper, Italian Beef, and Canadian Bacon.
Large Set Adds: Spinach, Cabbage, Eggplant, Lettuce, Peas, Cucumbers, Carrot (singular), Asparagus, Tomatoes, Beans, Squash, Okra, Collards, Potatoes, Cover Crops, Melons, Cauliflower, Mushrooms, Leek, Brussel Sprouts, Rhubarb, Ham, Onions, Anchovies, Black Olives, Mushroom, tomato (again, but singular), Pineapple, Bacon, Jalapenos (without a tilde), Potato, Italian Sausage, chicken, Green Peppers, garlic, salt, Parsley, Extra cheese and Zucchini. (Also, a few things from the first list were now in lowercase! And interestingly, Italian Beef was missing from the Large Set.)
Comment: What's with all the veggies? It didn't take long for it to turn Californian on me. Interesting about the Italian Beef.

Seed: AltaVista, Excite, Infoseek
Small Set Adds: Lycos, Yahoo, HotBot, Webcrawler, GoMuzi, Google (tisk tisk, 9th place!), Magellan, Northern Light and MetaCrawler.
Large Set Adds: Alta Vista (with a space), WhoWhere, Looksmart, ASK JEEVES (in allcaps, just like that), DogPile, Msn, Snap, Direct Hit, Netscape, GoTo, Web Crawler, Starting Point, Galaxy, Goto com (no dot), search com, Hot Bot, Netscape Search, AOL Search, Open Directory, Open Text, AOL NetFind, DejaNews (still?), Iwon, Deja News (st ill?), FAST Search, Northernlight, AOL, Go com, Mamma, Inktomi, Savvy Search, MSN Search, Go and LinkStar.
Comment: Yahoo, Google, HotBot were each allcaps for some bizarre reason.

Just one more, please? I'm hooked!

Seed: Slashdot, Everything, Sourceforge
Small Set Adds: Freshmeat, NewsForge, Themes org, thinkgeek, Freshmeat net, Advogato, Kuro5hin, Linux com (hmmm), WikiWikiWeb (huh) and Debian (interesting...)
Large Set Adds: Ars Technica, Solaris Guide, Gnu, OSDN, The Register and... Webmonkey.
Comment: None I can think of. Ook.

The other cool Google Lab experiments, as of August 9th, 2003: Google Glossary, Voice Search, (Google) Keyboard Shortcuts, Google Viewer, Google Webquotes, Google Compute, and hey wow, as of August 4th: Google News Alerts!

Log in or registerto write something here or to contact authors.