Virtually everyone gets frustrated with Internet search engines. The most common complaint about them is poor signal-to-noise ratio. While much of this can be attributed to the average user's unfamiliarity with Boolean syntax and advanced filtering methods, it is also fair to say that the prioritization and ranking schemes used by most engines are equally at fault. Google, generally considered to provide the highest quality search results, uses an academic method of ranking citations, giving highest placement to those sites who are referred to the most by other sites. While this is better than most, what is still missing is a deeper level of qualification than generic numerical popularity.

Two alternative search methods, semantic engines and peer-to-peer (P2P) networks, currently represent what could be considered Internet search's "next generation."

Popularized by the recent Napster litigation, peer-to-peer Internet searching differs from today's search engines in that it requires a community of host computers to form a network on which searches are based, whereas traditional engines rely on submissions, human editors and robot "spiders" to search the Internet and compile listings.

How would P2P search work? One method, being pioneered by Project Pandango, would take a person's keyword search and examine it against the Web histories and bookmarks of an initial network of 100 referrers. From there, the application would search the Web histories of those 100 referrers' combined 10,000 referrers, and then a third cycle -- so that the query would canvass the Web pages visited and bookmarked by 1 million people.

Over time, it is hoped that enough people add their histories to the network, providing a mass of intelligence around which websites are popular among focused communities of like interests. For example, if you are a mechanical engineer in an aerospace company, you'd query the P2P network of people with the same profile, accessing the links deemed most popular by the entire network.

Critics of P2P note that it will take some effort to break user inertia, and that most will initially be reluctant to download and install the necessary software. But, as Napster demonstrates, it is certainly achievable, but requires critical mass before becoming effective.

The second method, semantic processing, promises more accurate search results based on context. The best example of this is Cobrain.com, whose methods are based on the Russian problem solving technique known as TRIZ/TIPS, which was originally used as a way to search patent databases for reusable engineering solutions. TRIZ software vendor, Invention Machine, ported TRIZ approaches to their Cobrain.com website, which claims the ability to search any databases of information with contextual intelligence. The current site is limited to searches on technical, medical or patent-based problem solving, but Cobrain asserts the technology works with any type of knowledge. One nice feature is that search results are accompanied by a pop-up window highlighting the location of the specific text within the site that matches your search terms, a feature that would be invaluable in current search sites. Two other companies, iPhrase and Verilytics, are also working on search tools that use advanced linguistic theories.

What impact might search evolution have on product development? In truth, the concepts behind P2P networks are not new, and have been employed for several years by some CAD and PDM systems as a file management tool. However, these previous applications may have been a little before their time. Renewed interest in P2P has surfaced in part because the Internet more easily allows the connection of distributed computers. Given time to develop, it is realistic to expect P2P networks to become more frequently applied within and between corporations for archive retrieval, cross-project knowledge transfer and reuse, supply chain collaboration, and much more. Decentralization of networks in general will greatly increase the amount of available content, creating an even higher need for accurate search capability.

In contrast, semantic searching is very immature, with Cobrain being one of the few useable sites employing these methods, but they promise much more accuracy and relevancy to information searches. Whereas P2P capitalizes on information gleaned from human behavior, semantic search engines such as Cobrain rely on artificial intelligence. XML makes similar promises, however, XML tagging is still subject to the oddities of human language and do not offer semantic or contextual analysis.

Bottom line, current Internet search engines are adequate, but only scratch the surface in terms of how useful they could become. As the Internet continues to grow in both number of users and content, it's true growth and usefulness will be determined by the limitations of these and other navigation tools.

Related Links:

• The Search Center - www.tka.co.uk/search/home.htm
• Project Pandango - www.pandango.com
• Cobrain - www.cobrain.com
• IPhrase - www.iphrase.com
• Verilytics - www.verilytics.com

NOTE: This article was produced in conjunction with a recent survey conducted by Management Roundtable and New Product Dynamics on the subject of "How product developers use the Internet". Two articles are available discussing the results of this survey:

• "How Product Developers Use the Internet" - April 2001, Product Development Best Practices Report - to download a .pdf reprint of this article: http://www.roundtable.com/PDBPR_Internet.pdf

We share reader reactions to TCP articles on our website.
Please send any feedback to gregg@roundtable.com

---

Interested in sponsoring this newsletter?
For a list of terms and rates, send an e-mail to gregg@roundtable.com or click here.

article-two:
Online Time Killers for Product Developers

The following two websites are presented by Soda, which I can only describe as a product development "creativity" firm located in the UK. While their main business is helping companies create unique concepts and implementation strategies, they like to play a lot too, and have created the following two sites to help you play with them. Note: make sure your web browser likes Java before surfing these pages.

Link #2: http://www.soda.co.uk/soda/loyaltoy

This page features a "customer relationship management game". If you've ever played an online Java or Flash game, this one will feel familiar. Here's how the game works: You are a CRM person at a travel agency. Floating around the screen are little guys that look similar to the old Pacman ghosts; these represent your customers. At the bottom of the screen, little icons scroll by that say "bookings" or "hotel reservations"; these are your products. When you click on a customer, he will tell you what kind of product he needs, at which point you just drag and drop the correct product icon on top of the customer. If you've done well, customers will smile loyally and help you convert new customers. Also lurking around are your competitors, who will try to lure your customers away. The game starts off a little slow, but eventually becomes a furiously paced exercise as you try to satisfy a screen full of roaming customers. This one is not as cool as sodaplay, but worth a few smirks.

article-three:
Top Ten Signs Steve Jobs Just Doesn't Care Anymore
...from the MRT satellite office in Cupertino, CA