Agents of Common Senseby Thomas CalwellThe flood of information swirling through the Internet and pouring out of our PCs has risen well beyond our fingertips - we're now drowning in the stuff. No matter how many search engines we might bookmark, when we rely on keyword searches we can spend as much time online wading through the data deluge as we would scooting off to the local library to find the answer. And as more information flows online daily, it's only getting worse. Engineers at Toshiba's Research and Development Center in Kawasaki are taking a novel approach in trying to find a cure for data diarrhea. They have come up with a way of creating and locating pertinent information by applying common sense to the problem. Literally.
Knowledge bases The system is only as good as the information that users put into it, so it has been deployed on an intranet to encourage wide participation in building up the knowledge base. According to Yoichi Takebayashi, a senior manager at the R&D center, some 300 researchers and clerical staff are helping in the task. All that is needed to participate is a browser and a useful piece of information. A user simply supplies a title for each nugget of knowledge, then keys in the information. Before sorting and storing the data, the system analyzes the content and extracts keywords that are used to flag the content for retrieval. The personal knowledge base is linked to a second database - the "office knowledge base" - that contains more systematized information about office procedures, job flows, documentation, and the like. In this case, specific Toshiba engineers are assigned to analyze how the center goes about its business and input the results, as in the case of the personal knowledge base. The result of all this is that a user can ask such questions as "My printer is jammed. What should I do?" or "What measures should I take in case of an earthquake?" or even "What forms do I need to fill out to get additional funding?" A software agent analyzes and interprets the query using natural language algorithms, then retrieves data pertinent to what the user is looking for. If the information isn't yet available online, the system will suggest the best person to contact. With questions concerning documentation, it can also call up electronic copies of the appropriate forms.
Developing an online mentor Neither is the system a form of FAQ (frequently asked questions). FAQs on the Internet are becoming as long and tedious to plow through as the download dumps from search engines. "FAQs can be huge," says Takebayashi. "Here, we are providing specific information on demand." As a manager, he admits he can rarely find time to individually mentor each of the young researchers under him and answer all their questions. But with this electronic agent system, he is gradually making his experience available to everyone who can benefit from it. While the system is easy to use, Takebayashi says the biggest hurdle so far is getting certain researchers to contribute what they know. "I input [to the knowledge bases] regularly," he says. "But it is difficult to motivate others to do the same. Researchers are more interested in conducting elegant research."
Better with time In time, the agent system will be upgraded to monitor each employee's computer, automatically extracting useful information as they tap away on their keyboards. In this way, users will be creating their own personal and office knowledge bases as they work, with no extra effort. Now that's an uncommonly practical way of using common sense.
John Boyd writes commonly sensible articles about technology for a number of publications, and hacks out the weekly Computer Corner column in the Japan Times. If you have any common sense yourself, you will not risk using fail mail to contact him at 6840615@mcimail.com. Back to the table of contents |