CREATING WAYS TO EXTRACT `GOLD' FROM INFO TORRENT

Author: By Peter J. Howe, GLOBE STAFF Date: 04/16/2001 Page: C2 Section: Business

If you work at a job that involves the Internet, you probably are less than totally satisfied with Net search engines or database-analysis software. They may get you answers, but only after a depressing amount of effort, and maybe well past the point when the information was most useful.

Two area companies, iPhrase in Cambridge and Verilytics in Waltham, are developing some highly advanced systems aimed at panning out the nuggets of data gold you want from the rushing rivers of information that surround us now.

Both companies are applying advanced theories of linguistic analysis to convert a question you ask into a set of facts extractable from vast sources of data - and an answer that can be sent back to you as an e-mail or a message on your wireless pager or phone.

The potential applications are dizzying, but today iPhrase is set to announce a fairly straightforward one: a simpler way to find a computer that meets your specifications. Cnet.com, the online news and comparison-shopping site that reported 27 million users last quarter, is rolling out an iPhrase-powered system this spring that lets you go to the Web site and ask, in one step, for a list of all laptop computers costing under $1,500, weighing under three pounds, with such-and-such processing power, available in black.

iPhrase is also working with discount broker Charles Schwab to launch a system that would let Schwab customers use its Web site to track down stocks meeting specific criteria, or get answers to questions about how different accounts work.

Verilytics is looking at the market for people like portfolio managers who are trying to find meaningful information from a torrent of stock-market data and news. One might be: Page me any time IBM stock moves 5 percent more than the market, its chat-board traffic is triple normal levels, and news wires are reporting an earnings surprise at any of 10 IBM competitors.

"The problem of information access has largely been solved," says Verilytics founder and CEO Shikhar Gosh. "But if you say there's going to be infinite data, that's an insoluble problem. How do I sort through all this stuff and get to the two or three things that are important now?"

Through its language analysis, Verilytics can link share-price movements to reports about "legal problems," "analyst comments" from Wall Street movers and shakers, earnings announcements, or "buzz" levels at chat boards.

Two-year-old iPhrase, whose founders include three MIT Laboratory for Computer Science veterans, offers a similar tool that basically plays back the question it understood you were asking, with the answer presented on the screen below.

The kinds of questions you can pose include, "Do you have any minivans for under $32,000 with seven seats and side airbags?" Or "Which biotechnology companies have the most cash on hand, and where are they located?"

These technologies are far from cheap. A big Web site or Fortune 500 company could expect to spend hundreds of thousands or millions of dollars adding the systems to make extracting useful, specific information simpler. But both iPhrase and Verilytics are optimistic corporations will conclude that incremental spending to help find needles in data haystacks will make irresistibly good business sense.

Among those giving the Verilytics system a test drive is Household Credit Services of Salinas, Calif., the ninth biggest US credit card issuer, whose clients include General Motors and the AFL-CIO.

Household has used a Verilytics-produced Web site since the company's days as iBelong.com. Now the 5,000-person company is evaluating a Verilytics system that monitors millions of credit card transactions and other financial data to find out when a card may have been stolen, or when a subscriber with credit trouble might warrant a pager message sent to a credit manager.

"In this industry, data is our lifeblood," said Michael Marcus, managing director of new ventures for Household. "The business model they're focused on is one where there are not a lot of good solutions out there today."

Peter Howe can be reached by e-mail at howe@globe.com.