Cutting through search-engine clutter

With over 8 billion Web pages on file, Google is the most comprehensive index of the Internet. Too bad.

Internet users don't need 8 billion pages, or 8 million. Even eight is usually too many. More often than not, we'd settle for just one page -- the one with the information we need.

Google understands this. Indeed, its early success was due to page ranking, a clever statistical technique that does a pretty good job of guessing what the searcher is looking for. But Google's best gimmicks still fall short, and we're often compelled to slosh through waist-high data to find the facts we seek. So despite Google's spectacular success, there's plenty of pent-up demand for something better, and many an entrepreneur hoping to satisfy it.

A couple of the latest entrants come from the other side of the pond, Britain and Israel, while a third hails from Pittsburgh. None of them are search engines in the traditional sense. Instead, these new entrants let others index the information, while they figure out better ways to pick out just the bits that matter.

Previewseek.com, which opened for business earlier this month, does this the hard way, with lots of fancy math. The London-based firm was founded by Harvard graduate Christopher Hong, who bluntly declares that Previewseek ''will beat Google as the world's most advanced search engine."

But he who shoots at the king must kill him, and Hong's aim isn't that good. In its present form, Previewseek has a certain appeal, but it's hardly a Google-killer.

Rather than create its own index, as Google does, Previewseek searches several online indexes, such as Microsoft Corp.'s MSN Search, and combines the results. Big deal -- such ''metasearch" services have been around for years. But Previewseek goes further, by using home-brewed software to pore over the results and sort them into meaningful subcategories, so the user can home in on the most useful information.

Type ''Toyota" into Previewseek. You'll get the usual pageful of Web links, and considerably more. In a box at the top of the page, you'll get a brief history of the Japanese automaker, culled mostly from the Wikipedia online encyclopedia. Along the left side of the browser are links to pages that contain words related to Toyota -- words like ''truck," ''auto parts," and ''dealer." A searcher looking for information on various Toyota products can quickly assemble lots of relevant information. The site even lets the user instantly save up to 30 search results in a special folder, for later review.

It's not half-bad -- more like 25 percent bad. Blame it on the software that searches out related pages. It works quite well for some topics, such as Toyota. But many searches produce bizarre results. On a whim, I tried typing ''website banner." Up came a lot of appropriate pages about Web design, mixed with others devoted to the breeding of cats. Heaven only knows why. Even Hong admitted that Previewseek's software needs a major upgrade.

But we needn't wait. Instead, we can turn to Clusty.com, a search service developed by the computer science wizards at Carnegie Mellon University in Pittsburgh. It's Previewseek done right -- or at least a whole lot better. Like Previewseek, Clusty sorts search results into related categories, or clusters. And like Previewseek, some of these related clusters make no sense. Clusty also associates website banners and cat breeding, for instance. But Clusty seems to deliver far fewer illogical links and a much higher percentage of worthwhile data.

There's a good chance you've never heard of Clusty, but there's also a decent chance you've used it. Company founder Raul Valdes-Perez estimates that 10 percent of Americans have used Clusty. That's mainly because the America Online Internet service, with over 20 million subscribers, displays Clusty results with every search.

Despite their flaws, Clusty and Previewseek make it easier to bypass worthless information and close in quickly on the knowledge you seek. But the process was even easier in the days when people did their research in libraries. Bookworms learned to rely on reference books where they could find vital facts fast, with a minimum of distraction.

At last, there's an Internet tool with the same kind of power and simplicity. Answers.com, an Israeli company founded by Massachusetts Institute of Technology graduate Robert Rosenschein, is the best online destination for anybody in search of a few quick facts on nearly any subject.

Answers.com doesn't search the Web, and its gigabytes of irrelevant gabble. Instead, Rosenschein and his partners struck deals with over 100 reference databases, ranging from Wikipedia to Fogwell's Guide to Wine.

Yes, wine. Type the word ''claret" into Google, and you get 871,000 Web pages where the word appears. Type it into Answers.com, and Fogwell informs you that a claret is a red Bordeaux wine. You also get a definition of the word from the American Heritage Dictionary, a brief history of the wine from Wikipedia, and 14 translations of the word into different languages, including Arabic and Japanese. All this, collected onto a single Web page and available at a glance.

Answers.com makes no pretense of being a Google-killer; indeed the two companies are on excellent terms. Google uses Answers.com to provide word definitions to its users, while Answers.com displays Google search results for topics not covered in its databases.

Users with Microsoft Windows computers can look up Answers.com information while in the midst of almost any computing task. The company offers a free software download that lets you research an unfamiliar word or phrase simply by clicking on it. Up pops a browser window, with a page of answers from Answers. Just one page. It's all the information many Web surfers will ever need.

Hiawatha Bray can be reached at bray@globe.com.