Crib Sheets For Folks Swamped With Data
Electronic data bases place the world at your fingertips. Trouble is, even a 15-minute search can leave you digging for hours through the documents you have downloaded into your computer. The touted Information Superhighway could make things worse: Wayfarers are likely to get buried in a multimedia blizzard.
To help sift through the avalanche, researchers at Toshiba Corp. have created a text-retrieval system that they say can digest long documents and produce abstracts of them in simple, grammatical Japanese sentences. If the abstract is inadequate, a customer could tell the system to redo it and, for example, request more technical details or market statistics.
Toshiba researchers figure it will take two more years to commercialize the system, which runs on a Unix-based workstation. First, they must teach the system to search for complete sentences, rather than just key words. Second, they need to speed up document analysis, which can take hours of computing time. Finally, they need to convert the software to English, since the vast majority of the world's data-base customers reside in the U.S., not in Japan.