RESEARCH ON INTELLIGENT QUESTION-ANSWERING SYSTEMS

Abstract : This report describes progress toward an 'intelligent question- answering system'--a system that can accept facts, retrieve items from memory, and perform logical deductions necessary to answer questions. Two versions of such a system have been implemented, and the authors expect these to be the first in an evolving series of question-answerers. The first system, QA1, is based upon relational information organized in a list-structured memory. The data consist of general facts about relations as well as specific facts about objects. QA1 has limited deductive ability. QA2 is based upon formal theorem- proving techniques. Facts are represented by statements in the predicate calculus. Although the memory organization is simpler than that of QA1, the sophisticated logical abilities of QA2 result in greater question-answering power. The report gives examples of the performance of QA1 and QA2 on typical problems that have been done by previous question-answerers, and describes plans for extending the capabilities of QA2.