NATURAL LANGUAGE LEARNING BY COMPUTER
CARNEGIE-MELLON UNIV PITTSBURGH PA DEPT OF COMPUTER SCIENCE
Pagination or Media Count:
Learning a natural language is taken as an improvement in a systems ability to express situations in a natural language. This dissertation describes a computer program, called Zbie, written in IPL-V, which accepts the description of situations in a uniform, structured functional language and tries to express these situations in a natural language. Examples are given for German and, mostly, Russian. At run-time, Zbie builds simple memory structures. Patterns and sets are built on the functional language. The translation rules of the patterns and an in-context vocabulary provide the transition to the natural language. Zbie is a cautious learner, and avoids errors by several mechanisms. Zbie is capable of some evolutionary learning.
- Computer Programming and Software