AD/A-005 04t COMPUTER GENERATION OF NATURAL LANGUAGE FROM A DE E P C ONC E PTUAL BASE

For many tasks involving communication between huivans and computers it is necessary for the machine to produce as well as understand natural language. We describe an implemented system which generates English sentences from Conceptual Dependency networks, which are unambiguDus. language-free representations of meaning. The system is designed to be task independent and thus capable of providing the lar.gnage generation mechanism for such diverse problem areas as question answering, machine translation, and interviewing. A dissertation submitted to the Department of Computer Science and the Committee on Graduate Studies of Stanford University in partial fulfillment of the requirements for the degree of Doctor of Philosophy. This research was supported by the Advancul Rtscarch Projects Agency of the Department of Defense under Contiact DAHC l}-7}CO-f35 and the National Science Foundation under Contract NSF OH206 . The views and conclusions contained in this document are those of the author(s) and should not be intapre'ed as necessarily representing the official policies, either expressed or implied, of Stanford University. ARPA. nsf. or the U. S. Government Reproduced in the U.S.A. Available from the National Technical Information Service, Springfield. Vitfrtr2/5/