The MikroKARAT Distributed

Knowledge Acquisition Environment

Dr. Ralf D. Brown

Carnegie Mellon University

Center for Machine Translation

5000 Forbes Avenue

Pittsburgh, PA 15213-3890

Abstract

MikroKARAT (Mikrokosmos Knowledge Acquisition and Review Assistance Tools) is a system written in C++ with the Motif user interface which permits multiple geographically dispersed knowledge acquirers to simultaneously update a large knowledge base. It uses a client-server model, with each acquirer running a client environment which communicates over the Information Superhighway with a central server containing the permanent repository of the knowledge base and providing arbitration between users.

The knowledge bases consist of a set of linked frames, and the central knowledge-entering component is a flexible structured editor for manipulating these frames. Depending on the type of knowledge base, it can be configured for frames with anything from very rigid formats to virtually free-form contents. The editor proper is called, with the appropriate configuration, by various components specialized to a particular type of knowledge base, such as lexicons or ontologies.

Other applications in the Mikrokosmos project will also make use of the underlying frame representation package FramepaC used by MikroKARAT, which provides essentially transparent access to frames stored in either a file or remotely on a server as though they resided in memory. In this way, an MT system can continue using a knowledge base even as it is simultaneously being enhanced.

Body of Paper


Images
Tables

Ralf Brown (ralf+@cmu.edu)