Toshiba Develops an Electronic Agent System for Knowledge Sharing

4 February, 1997

TOKYO--Toshiba Corporation today announced the development of an electronic agent system that supports efficient sharing and reuse of business knowledge among individuals in an organization. The agent system analyzes, interprets and accumulates information input to a database by individual users, and retrieves required information whenever a user enters a query in natural language via its intelligent human-computer interface. Its automatic knowledge structuring and intelligent retrieval facilities enhance productivity and promote mutual understanding among individuals.

The electronic agent system was developed at Toshiba R&D center, where it is currently being evaluated and improved through practical use. It consists of
1) a knowledge base which stores business-related information; 2) a human interface that interprets natural language dialogues with users, extracts information needed from them, and retrieve the information from the knowledge database; and 3) a multimedia information structuring facility that analyzes information input by users and structurally registers it in the knowledge base.

The electronic agent system for knowledge sharing can be accessed within the intranet of Toshiba's R&D Center, where authorized users both register and retrieve work-related information and advice over their personal computers. Currently, the knowledge in the database includes information on intellectual property rights, employee information, administration and other business matters.

While advances in computing promote the gathering and sharing of huge volumes of data, the very volume of information and know-how available makes it difficult to assure that people have the information they want when they want and need it. Searches based on using key words are becoming inefficient as the amount of data continues to increase. Toshiba's new information sharing system uses a natural language dialogue interface in order to make it much easier to search for necessary information.

Another problem is the lack of a practical means to integrate databases of individual business know-how. As experienced workers are often too busy to share their expertise, useful information that should be shared goes to waste. The new information sharing system allows easy integration of individual know-how, which the system can then provide in place of the information owners.

The information and knowledge sharing system will be demonstrated at Tomorrow 21, an exhibition of Toshiba technologies, at Tokyo International Forum from March 4 to 9.

Three technologies realize information sharing system:

    1) Office Knowledge Database and Know-how Database
    The electronic agent system has know-how database and office knowledge database. The know-how database includes a diverse range of useful data, such as advice on writing good technical papers, know-how on announcing technical achievements, advice on operating computers, patent-related problems, and information on office security. The office knowledge base contains systematized data on the office and business knowledge. This includes office procedures, job flows and documentation, relations among persons and business groups. The new system connects the two databases and offer more flexible and accurate searches for business information than any other database system.

    2) Natural Language Dialogue Interface
    Toshiba's natural language processing technology allows the system to understand intentions contained in written Japanese sentence(s), enabling intelligent searches for required data among a large volume of data. A user can ask natural questions, such as: "What documents are required when purchasing office furniture?" or "Please tell me who is familiar with edutainment software?"
    Required information can be retrieved more easily and efficiently than with systems based simply on key-word searches, and with much less unnecessary information inevitably turned up in key-word searches. In addition to text, support for multimodal interactions can be added to the system. This allows the agent to interpret utterances and gestures, creating an environment supporting natural dialogues.

    3) Automatic Structuring of Multimedia Information:
    Quick and efficient retrieval of data from databases is assured by the agent's ability to automatically analyze and identify data input by users. It extracts key words from input data, attaches identification tags and structurally registers the data in the knowledge database. Multimedia information is also registered as know-how data, using Toshiba's newly developed multimedia converting technology.

System Configuration:

Information in the press releases, including product prices and specifications, content of services and contact information, is current on the date of the press announcement,but is subject to change without prior notice.