UNITRAN: An Interlingual Machine Translation System.
MASSACHUSETTS INST OF TECH CAMBRIDGE ARTIFICIAL INTELLIGENCE LAB
Pagination or Media Count:
Machine translation has been a particularly difficult problem in the area of Natural Language Processing for over two decades. Early approaches to translation failed, partly because interaction effects of complex phenomena made translation appear to be unmanageable. Later approaches to the problem have been more successful but are based on many language-specific rules of a context-free nature. To try to capture all of the phenomena allowed in natural languages, context-free rule-based systems require an overwhelming number of rules thus, such translation systems either have limited linguistic coverage, or they have poor performance due to formidable grammar size. This report presents an implementation of an alternative approach to natural language translation. The UNITRAN Universal Translator system relies on principle-based descriptions of grammar rather than rule-oriented descriptions. The approach taken is interlingual, i.e., the model is based on universal principles that hold across all languages the distinctions among languages are then handled by settings of parameters associated with the universal principles. The grammar is viewed as a modular system of principles rather than a large set of ad hoc language-specific rules. Interaction effects of linguistic principles are handled by the system so that the programmer does not need to specifically spell out the details of rule applications. Only a small set of principles covers all languages thus, the unmanageable grammar size of alternative approaches is no longer a problem.