The Enterprise Approach to Search
A French search company believes the key to better Web-based search comes from the corporate world.
Exalead, a Paris-based company, is experimenting with a new search model. It believes that by combining the approach of “enterprise” search and Google-like page-ranking search, individuals will be able to surf the ever-expanding Web more thoroughly and efficiently.
Enterprise search, which scours the intranets of institutions, has been the previous focus of Exalead’s products. It differs from Google’s page-ranking Web search approach: Google’s tack is to give a higher rank to Web pages with more incoming links, explains Dan Gruhl, research staff member at IBM Almaden Research Center in San Jose, CA. This approach has created a “dark” Web, however, in which a large amount of data goes unread. Companies don’t have the luxury of having all of their documents ranked, Gruhl says, and therefore page ranking algorithms “work pretty poorly” in the enterprise search environment.
In contrast, enterprise search software finds relevant items by processing an entire document or file “as a concept,” Gruhl says. It’s able to extract the meaning of a document, instead of just scanning it for key words. This allows the engine to search for terms in documents that weren’t even entered by the user. For instance, if an employee at a pharmaceutical company searches for “aspirin,” the search software would extend the search to “pain killers” and “COX inhibitors” as well.
This is the crux of Exalead’s blending of a Web-based and an enterprise search approach, with the entire set of search features needed to find a specific file applied at the user-search level.
Using Exalead, an individual has a variety of options: searching by phonetic spelling, by six different languages, by specific document type (.pdf or .doc are two examples), and by date, continent, or country.
While many of these features are similar to Google’s “Advanced Search option,” Exalead also offers a “related terms section,” in case the initial search terms don’t bring up the intended results. This enables what François Bourdoncle, the company’s CEO, calls “serendipitous search,” in which a person can find something even when he or she doesn’t ask the right question.
These related terms help distinguish Exalead from Google’s advanced search, by allowing users to start at one search query and progressively narrow it down. For instance, a search on “greenhouse effect” could yield the related term “greenhouse gases.” From there, a user can further narrow the search to geography, or types of document, for example. Then, one could return to the initial search and narrow down again, using a different related term.
Still, searching the Web effectively, even with this hybrid approach, is not technologically easy, Bourdoncle says, and it has taken his team of engineers – many of whom came from Alta Vista, the Internet’s first viable search engine – about eight years to put in place.
The largest issue for any company to effectively search the Web is scalability. Large institutions can have up to 100 million documents that need to be sifted through by a search engine, but the Internet is composed of billions of pages. Bourdoncle explains that Exalead has broken up the Internet into several indexes that share the task of crawling the pages and doling out search queries. The software is designed to work best with Exalead’s index structure, and, Bourdoncle says, was written to be able to grow with the Internet. Currently, the company indexes 4 billion pages, but it hopes to have 8 billion – close to Google’s capacity – by the end of the summer.
All of these features add up to a search engine that has a very different feel from Google – not surprising, since Google was designed to search Web, while Exalead was designed to search company servers.
But looks matter too. Google has a simple appearance, it’s easy to use, and it’s familiar. “The user interface is always important,” says Ray Larson, professor of computer science at the University of California, Berkeley. He says Google’s single search box seems to be the preferred choice. And while more options in a search allow a person to narrow down the results, the first 10 hits that Google throws up may be good enough for many searches.
Bourdoncle is aware of the challenges his company faces with usability. Admittedly, Exalead takes some getting used to, and can be frustrating if you don’t want all the detail that it can offer. They’ve conducted user trials, Bourdoncle says, that have helped it decide how to simplify the search results screen and hide some the search options (although these tools are only a click away if they are needed).
So how will an obscure French search company make its way in the world of Google and Microsoft and enterprise search? “Our plan in a few years is to be in the top five Web-search engines,” says Bourdoncle, eventually selling advertising, as Google does.
Until then, enterprise search solutions are still the company’s principle money maker. In fact, Exalead offers an integrated search package that includes Web, enterprise, and desktop search (a version of the desktop search is available as a download). And in this arena, Bourdoncle says, the company has more lofty goals: to be one of the top three enterprise search companies, behind Microsoft and Google. “We’re positioning ourselves as the challenger,” he says.
It remains to be seen how effectively Exalead will combine enterprise and Web-search tactics – and how well the public will respond. However, Berkeley’s Larson says that an enterprise approach could translate well to the Web. It will depend on how the company deals with a larger and larger Internet. “If you can scale up gracefully,” he says, “you can handle almost any search problem.”