Search engines from Google, Yahoo and Microsoft may do the trick for users doing keyword searches for restaurants, maps or other bits of information. But what about professional researchers looking to access content buried in the world's billions of Web pages?
These folks have been largely out of luck, until today. Startup Infovell Sept. 8 launched its "research engine" at Demo 08 in San Diego, Calif., to help workers in medical, biopharmaceutical, patent and even business sift through the so-called deep Web to find articles and other valuable information from industry experts.
I met with Infovell CEO William Park a few weeks ago and the technology is impressive. Infovell uses a patented algorithm called KeyPhrases. KeyPhrases searches phrases up to 25,000 characters per query to help searchers extract more meaning and relevance from documents in the research engine's index of 500 million-plus Web pages.
Infovell indexes medical journals from a network of biopharmaceutical and life sciences content providers to provide researchers access to important content.
Users don't have to be super insightful with their queries; they may cut and paste a paragraph, or even an entire article and click the More Like This button to find all related results. Moreover, it searches queries in any language, which means the research engine is immediately viable for use worldwide.
Unlike Google and others, which rank pages to prioritize results based on popular searches, Infovell's research engine ranks results purely on how closely the content matches what the user is looking for. Often, the bigger query, the better to extract greater meaning and relevance from documents. Park told me:
"Search engines are too general for someone trying to go deep on a subject. Professional researchers don't want a quick answer, they want a comprehensive answer from the deep or hidden web and 99.8 percent of the information on the Web is not accessible because information is subscription protected on research sites, such as medical journals, or it's in databases."
What I was most struck by was the graphics with which Infovell renders search results. The research engine uses a clustering technology that will group search results by title, author, date range, keyword types and data source in a Venn-diagram-like cluster tree.
"We will look at all of the words and phrases in the results and cluster them around statistically like-minded phrases and words," Park told me.
Users may then quickly sort through results and tag search queries and results for easy recall. In wiki-like fashion, users may also organize their results into visual folders or set up alerts that will tip them off to new findings.
Infovell's research engine will be available beginning September 22 as a hosted, premium service for individuals, universities and businesses. The company is offering an early adopter $45 per user, per month subscription.
This will eventually balloon to $200 per month per user but the price per user will scale down for organizations who buy a license for the whole Web site.
Park told me Infovell will release a free version of its research engine on a limited basis for consumers who want to search the Deep Web but don't have the need for some of the advanced features available in the premium version.