Coreintellect, Inc. (http://www.coreintellect.com), a Dallas-based advanced-search technology company, has launched a flagship product called Core360, which applies a full array of post-Boolean search engine and data-handling techniques to familiar full-text and directory business data. According to Perry Copus, Coreintellect's CEO, the new product was designed to solve the two major problems of business searchers: data overload and data overlooked.
Coreintellect's technologies include word-sense disambiguation, context-sensitive queries, real-time content filtering, information extraction, and concept-based user profiling. Marketing of the new service is targeted toward two audiences: individual users paying a low monthly fee and Fortune 2000 company users with institutional licenses. At first glance, the new Core360 portal service seemed to have a number of areas that needed tweaking, but that's hardly unexpected for a brand-new service. However, the data extraction of information based on natural language queries in the Quickanswer feature (which Copus plans to rename Coreanswer) marks Core360 as a service to watch.
Core360 includes the following content suppliers:
"There are a number of other deals we're inking—e.g., SEC EDGAR. We're working on … a variety of other deals in technology and business publications, government data, and additional market research. Right now we have over 1,000 sources," said Copus.
- U.S. and international news wires, such as Reuters
- a COMTEX feed (the full range of its content)
- 700 full-text titles from Bell & Howell (150 newspapers and business/industry weeklies and monthlies, particularly those focused on technology and trade)
- Thomson Financial's public company profiles, broker/analyst research (Investext), and market research from a number of top-tier firms
- CMP Media, including its publications Internet Week, Information Week, Tele.Com, and Network Computing
- Gale Group, which supplies 150-200 publications (including approximately 50 technology/trade journals as well as some in other areas), and databases (which will be integrated in the next few weeks, like Companies and Their Histories)
- Dun & Bradstreet private and public profiles for 6 million companies
Coreintellect is licensing the data according to fixed rates. Copus said the company feels this approach gives it more scalability options. Tim Levey, Coreintellect's chief financial officer, indicated that the organization was particularly interested in expanding data acquisition from the federal government. It already has such sources as the CIA World Factbook, but is looking to acquire more data from the U.S. Department of Commerce.
Coreintellect is looking to acquire Web-only content for Core 360. According to Copus, "[A]s we speak, we're integrating data from a variety of high-value sites for business information as well as doing some spidering. The richest and most valuable content is business-related, but we cover some general Web content as well."
The Core360 intelligent agent technology array uses a number of techniques and applies them to specific sectors of the problem. Copus described this approach: "When I started this company, I started by looking at the business problem, not the technology. That approach gives you a huge amount of freedom to bring together solutions. You can't solve all the problems with statistical techniques alone, not with pure information extraction, [and] not with semantics in broad text domains—that only works in narrow domains for real work. None of them are the silver bullet by itself. You have to take parts of each that make sense for different parts of the problem and leverage the benefits of all of them. Use each where each makes sense. That's the benefit of approaching the problem with a clean piece of paper. Our solution set wasn't constrained."
In the past, some of the services that have tried to apply post-Boolean search and data-handling techniques to real-world data services have found themselves struggling when the data or the user flows got too large. Copus recognizes the problem of scalability but doesn't fear it. "Information retrieval and statistical techniques are very scalable," he said. "NLP [natural language processing] and deep analytic techniques are highly accurate and precise, but not scalable at all. Information extraction when you have a broad group of items from open-domain text content creates intractable problems. The foundation of our system is statistical and filtering, but we have injected extraction and NLP in certain parts of the processing without impacting scalability."
The Core360 system allows users to select from taxonomies of companies, industries, themes, and concepts. From these, users can create customizable profiles to support current awareness. Upon entering the system, users see titles from recent content additions reflecting their appropriately grouped interests. They can then click on the titles to see the full-text. Some of the full-text—in particular, high-priced market research reports—will have pay-per-view surcharges involved. The system will warn users when this occurs. Institutional users may have access to pay-per-view sources controlled by on-site administrators, a feature that will be supported by software improvements that are under development. Coreintellect also plans to add a relevancy feedback feature that will allow the system to track user behavior and improve the profiles automatically. That feature should be added in the first quarter of next year, according to Copus.
The Quickanswer (or Coreanswer) feature prompts the user for natural language questions, such as "Who bought company X?" and "How much did it pay for the acquisition?" The system then searches the full content of Core360's archives and comes back with extracted sentences from sources that may contain the answer. Each extract has an abbreviated citation attached to it. Users can click on the sentences to read the full report. I tested this feature by asking Core360 who owned Netscape. In a display of 16 extracts, only one gave me the answer (America Online).
Coreintellect plans to market the package to business users, particularly decision makers at Fortune 2000 companies. A direct sales force will target these companies, but the organization also hopes to track the domain names of individual users who come to its portal site. A subscription to Core360 costs $25 a month, a little less if you subscribe quarterly, even less if you subscribe annually. The subscription price includes searching and content, except in the case of pay-per-view material, such as market research reports. At present, even institutional users will have to go to the Core360 portal to access the service, though Coreintellect has intranet interface connections under development.
As for the future, Copus indicated that Coreintellect intends to concentrate on Core360 as its primary service. Next year it plans to introduce more vertical-market business-information portals using the same technology. However, Copus didn't reject notions of applying Coreintellect's software in other roles, such as internal corporate data management. He also plans to develop applications for WAP wireless devices. The data-extraction feature on Core360 accommodates the small, awkward screens commonly available on such devices.
Coreintellect is a privately held company built with investments from Allexxis Capital, Autem Technology Partners, Billing Concepts Corp., CMB Capital Corp., ComHoldings Ventures, and Sanchez Capital Partners.
All in all, Core360 seems to be an interesting new service. For more information, check out the Web site (http://www.core360.com) or call 877/497-6898.