Putting it Together: Taxonomy, Classification & Search

by Jeff Morris

Continued from [ page 3 ]

According to Shipp, DataStream is able to convert the entire Federal Register into a Web-enabled, easily searchable format each day in less than an hour. DataStream also processes the Congressional Record, U.S. House and Senate Committee Reports and Bills and the Congressional Research Service's Bill Digest, making the data available on the Internet before most people arrive at their desks in the morning.

DataStream's techniques for data validation differ from the traditional approach, says company president Mark Anstey. "With the huge volume of data we process, you can't have a human look at every line," says Anstey. "Our validation routines automatically check every aspect of the content, formatting and markup." And errors are common in the data his company receives. "We've found errors in laws and regulations, incorrect citations in legal opinions, and a variety of other human and machine-induced data anomalies. If it's not 100 percent correct, we'll find it."

One of the ways CQ is using metadata is in developing a new version of a Web site called CQ Homeland Security. The project entails pulling together all Federal Register documents dealing with homeland security issues. "Our Web site will pull out and present the latest homeland-related pending rules, meeting notices, etc.," Shipp explains. "DataStream does the XML conversion, and our system allows the documents to be grouped into the appropriate subjects." That grouping is essentially classifying all the documents into a taxonomy — one that significantly reduces the amount of search time that might otherwise be required.

Here's the Best Part

Sunnyvale, CA-based National Semiconductor Corp., a leading manufacturer of analog electronic system components, has provided online knowledge bases to support customers since 1996. To ease search and navigation, the company built a custom solution built on One Step search technology from San Mateo, CA-based iPhrase.

National implemented One Step to help customers reduce the time-to-market for new products by providing fast and easy navigation through National's extensive online product catalog. National's vice president of Web business, Phil Gibson, explains that the expert design engineers who are National's primary customers know the exact parameters they need for parts to fit the requirements of what they have to build.

Customers visiting National's www.Solutions.National .com site are presented with an easy-to-use, interactive online design support environment that speeds their search for the best products for a particular application. The site features 30 color-coded block diagrams organized for specific applications in eight major market categories: automotive, broadband communications, displays, industrial, medical systems, personal/consumer electronics and wireless communications. These categories provide a taxonomic breakdown of National's components that saves engineers from searching through countless irrelevant parts lists. Design engineers can use the graphical navigation guide to access the 30 system solutions, get a list of recommended integrated circuits, perform thermal or electrical simulations online and order product samples.

Gibson explains that the functional diagram shows what the system architecture looks like. Functional blocks within each diagram show specific components and offer parametric selection guides listing all the parameters the engineer specified. "The interface has, essentially, leaped from that diagram and presented and sorted columns of the best components in that category," explains Gibson. Listings may change from day to day as National changes its lineup of components.

The selection guide, Gibson notes, is completely user-configurable; engineers are able to adjust parameters such as supply voltage, package type, pin count, price, and more than 300 electrical characteristics. In so doing, they are relying on a taxonomy that groups products according to specific characteristics. While National built the diagram, its underlying architecture and the ability to instantly adjust product specifications are made possible by iPhrase One Step.

"Our old diagrams were a hard-wired list of parts," says Gibson. "Others had copied that idea. Now I can multiply the number of diagrams to provide yet another level of service. I maintain a parametric table, populate it with information and harvest directly from those tables to the infrastructure. We have system experts who tune the functional query to pull up the best possible answer. This is much less work and provides better results."

By combining taxonomy, classification and search, "we're giving users the ability to pull the precise needle from a haystack of information," concludes Gibson.

While Gibson could not provide any return on investment figures, the fact that 50,000 engineers a day visit the site is, at least, anecdotal evidence of its success.

Jeff Morris
is a freelance writer based in South Salem, NY. He may be reached at jpm55@earthlink.net.

[ BACK | NEXT ]