|
Best of the Bots Documents Delivered to Your Door By Don Barker
Documents can find you in one of two basic ways, either as a Community Directory or
Personal Wire. A Community Directory (or guide) provides documents based on group
interests. For instance, Lycos integrates WiseWires intelligent agent technology
within its infrastructure to offer a collection of Web Guides (Community Directories). Figure 1 shows the Space/Sci-Fi guide at Lycos. Each Web Guide leads to a prioritized list of links about a specific topic. As an example, the guide for Mystery Science Theater 3000 includes links to instructions on how to build bots. These bots aren't software agents but replicas of fictitious robots, like Tom Servo and Crow T. Robot, that appear on the TV show Mystery Science Theater 3000. These cute mechanical men help keep Mike Nelson, the hero of the show, sane by making witty remarks during the endless series of terrible science fiction movies Mike is forced to watch. The content and priority of information in the Web Guides at Lycos are based on community opinion. Visitors are encouraged to use a simple three button rating bar to indicate how well they liked a particular page in a guide. User preferences then are processed by WiseWire's unique collaborative-filtering and learning agents to shape and rank the content within these community directories. In addition to community directories, documents can find you through your own Personal Wires. To Personalize a Wire, first select a Community Wire (Directory). Although Lycos doesn't yet have Personal Wires, the technology is available at the WiseWire's Research & Development site. WiseWire also provides a search feature to help you locate Wires that closely match your interests. Once you locate a Community Wire with the type of content you want, say artificial
intelligence, click the Personalize Button in the upper right corner of the WiseWire uses a combination content-filters, collaborative-filters, and WiseWire receives millions of documents each day, including Web pages, newsgroup postings, and premium sources like Reuters, Associated Press, Business Wire, PRNews, SortsTicker, and much more. In the first stage (or layer), the system uses content-based filtering to recognize key conceptual material, including multimedia, and encode it. As shown in Figure 3, learning agents continually fine tune and monitor the system at every stage. The encoded representations from the first stage are passed to the prescreening layer, where documents that don't fall into areas of interest are sifted out. In the third stage, the remaining documents are distributed into broad categorizes. The fourth layer ranks each document based on group interest in a particular Wire. In the fifth stage, the stream of documents are filtered based on the interests of each individual within a specific area. The final layer in the Filter Server Network uses a collaborative filter to match highly rated documents with the people most interested in receiving them. WiseWire's amazing pattern recognition capabilities also play a crucial role in making it possible for the system to process the ever growing amounts of information on the Internet in real time. By predicting the value of new documents coming into the system, the collaborative filter vastly reduces the time it takes for individual documents to find interested users. According to Ken Lang, the founder of WiseWire and a doctoral candidate studying machine learning and artificial intelligence at Carnegie Mellon University, the "...biggest accomplishment we made in the transition from Carnegie Mellon to [a commercial site] was [developing] something that could scale up to say a hundred million people." With this kind of power and capabilities,
it's little wonder that all kinds of companies are rushing to license WiseWire technology.
Everyone from financial publications and pharmaceuticals to virtual communities and
retailers are signing up. So, if you are looking for a way to mine the "jewels" buried in the wastelands of cyberspace, maybe it's time to check out some wisdom on the wire. |
|
|
What would the Internet be like if interesting documents sought you out
instead of you hunting for them down? Intelligent agents have long held the promise of
reducing information overload on the Net by delivering personalized information based on
your preferences. However, the reality is that most applications have been limited to very
narrow topics, such as entertainment-oriented subjects like books, music, and movies.
WiseWire, from the corporation of the same name, is changing all that with system capable
of adapting to any topic and letting them find you.
