The Secret Agent Man

How does Jango work?

Jango was originally developed by a pair of professors, Oren Etzioni and Daniel Weld, in the Department of Computer Science and Engineering at the University of Washington. These two researchers work in the area of artificial intelligence and intelligent agents. As the founders of Netbot the inventors of Jango, they continue to act as consultants for Excite on Jango technology.

Jango technology has won a number awards, including:

• one of the top one hundred achievements in science and technology for the 1997 by Popular Science.
• "Best Use of Intelligent Technology" at the 1997 WebINNOVATION
• "New Innovators Award" at the Business Online 1997 Show.
• one of 6 recipients awarded a 1997 CommerceNet VIP Award winner for Very Innovative Practices on the Internet

According to Eric Zocher, the Engineering Manager at the Seattle office of Excite, the power of Jango comes from its adaptive, fast, and scalable architecture. Jango relies on a collection of "Information Adapters" to identify and retrieve relevant information from Web sites in order to answer consumer queries. An Information Adapter is written for each merchant site, or product information source (i.e., site with product reviews), using a proprietary and patent pending computer language.

This high-level language was designed by Dr. Bob Doorenbos, a graduate student of the professors, to navigate and search a Web site automatically, using native Internet protocols. It handles many of the background tasks necessary to locate and retrieve merchandise and product information from a site, thus, making the creation of new Information Adapters a relatively fast and easy job.

For existing shopping categories, like Computers & Software, Excite programmers can use a template from another merchant site in the category, however, constructing an Information Adapter for a new category requires a bit more coding. Since the high-level language they use takes care of the search and retrieval requirements of a new site, the majority of the time necessary to create a new shopping category goes into designing and building a Web search form that is well suited for the characteristics of the category. For example, to build a "Stock Brokers" category, programmers would have to learn about what characteristics were available and important to potential customers before building a search form.

Although Eric declined to give me specifics about the AI technologies powering Jango, the professors who invented Jango have a Web site at http://www.cs.washington.edu/research/projects/softbots/www/softbots.html/   that details their research efforts in this area. The Web site presents material on the "Softbot" project, which involves applying artificial intelligence techniques to build intelligent Internet assistants (software robots). In describing their technology, the professor say that a "…Softbot accepts goals in a high-level language, generates and executes plans to achieve these goals, and learns from its experience. Custom-built execution and sensing modules enable the [agent] to interact with the Web in real time."

Since this technology was the forerunner of Jango, I think it is safe to assume that the Information Adapters use a similar approach to locate, gather, and organize information. One key element of the Softbot technology is the automatic construction of metadescriptions of remote database servers. If Information Adapters use metadescriptions, it would help explain why they execute so quickly when retrieving product information – at least some, if not all, of the information is already present on the Excite server. This would eliminate the need to query the remote merchant servers every time a consumer requests information.

Eric did tell me that Information Adapters are discrete objects that mediate information between the merchant site and the consumer. When a consumer specifies a particular product configuration, the relevant Information Adapters "volunteer" information. For example, when I asked for a notebook computer, with a minimum processor speed and screen size, the Information Adapters for the various relevant merchant sites (e.g. Dell, Gateway, Compaq, IBM, etc.) worked together to provide the requested information.

In addition to being adaptive and fast, Jango is also highly scalable. The independent object-oriented architecture of the Information Adapters makes it possible for Excite to add an almost unlimited number of merchant sites and shopping categories. As a consequence, Jango will be able to grow to provide shopping assistance, with convenient and comprehensive access, to the millions of products available for sale at existing malls, stores, and commerce sites on the Internet.

Return