Computerworld - Imagine a search engine that doesn't deal in the present but watches the future. It doesn't delve through events that have happened; it gazes beyond the present to know what will happen and alert you to it. While PubSub Prospective Search by PubSub Concepts Inc. in New York doesn't go quite that far, the company has come up with a clever capability.
"Today's traditional search model is based upon retrospective search, meaning information which is already out there and available to one of the major search engines," says Bob Wyman, chief technology officer and PubSub co-founder. "Prospective Search looks for future information on a given subject."
In reality, though, it's far from crystal-ball gazing. It's like telling a search engine, "Let me know as soon as there's more information on x." Prospective Search stores users' queries and matches them against new information as it comes out.
"In the past, you browsed, whereas now you can gain access as new material appears," says Charlene Li, an analyst at Forrester Research Inc. "It requires a change in the way people look at information -- like having a concierge service for information you want to see that comes up in the future."
This notification service is made possible by PubSub's proprietary "matching engine," which can match published data against stored queries in real time. The speed of the matching engine, at more than 3 billion matches per second, provides the foundation for an Internet-scale search service.
Wyman was the product's primary inventor, and the key developer was Duncan Werner, vice president of engineering. Angel funding of $2.5 million was enough to get the project started. The product was developed with the realization that more and more Internet information will be published or pushed as feeds using XML standards such as RSS and Atom. Following the definition of the core algorithms, the matching engine was developed as a component in C++.
"The challenge became, what topics available as published feeds on the Internet would be of interest to users?" says Wyman. "It then became apparent that weblogs, or blogs, were growing rapidly and could provide published data on a large scale."
Prospective Search's matching service continuously reads more than 13 million blogs, over 50,000 newsgroups, all Securities and Exchange Commission EDGAR filings, press releases from major wire services, earthquake data from the U.S. Geological Survey, and Federal Aviation Administration airport delay information to instantly alert users whenever items of interest appear. More topics and sources are being added to broaden the appeal of the service.
Yet according to Marc Strohlein, a content software technology analyst at Outsell Inc., a Burlingame, Calif.-based research firm, all those data sources are still not quite enough. "PubSub is not yet as comprehensive as other services such as Intelliseek or Fast Search & Transfer, and users don't have the ability to select sources," says Strohlein. "But PubSub is a lot faster at finding stuff, and it is free compared to big fees for using some of these other services."
The product has been thoroughly tested in the real world. Prospective Search is available on PubSub's Web site as a free service and boasts more than 200,000 subscribers. Now the big test is how to turn it into a working business model.
"Once Prospective Search is established and has a critical mass of users, the service will be monetized through the insertion of ads into the subscription results," says Wyman. "Secondly, Prospective Search can be provided to media Web sites as a way to establish a much more relevant relationship with their users. 'Powered by PubSub' relationships will be monetized through a service fee or a share of relevant advertising revenue."
Robb is a Computerworld contributing writer in Los Angeles.
Read more about Applications in Computerworld's Applications Topic Center.
Free Insider GuideCopyright © 1994 - 2012 Computerworld Inc. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of Computerworld Inc. is prohibited. Computerworld and Computerworld.com and the respective logos are trademarks of International Data Group Inc.
