机器翻译词义辨识对策

词义辨识是机器翻译中最困难的问题之一,不夸大地说,它是全自动高质量机器翻译成败的关键。本文结合许多实例,详细探讨了解决这个问题的各种方法及其相互关系,这些方法包括成语加工、语法分析、语义分析、语境分析、代指分析、互补分析、中性词义以及一词多译等。它们的有机配合可以有效地解决绝大多数多义区分问题,从而为机译系统的实用化奠定了基础。所有这些方法都已在JFY-IV英汉机译系统中得到广泛运用和验证。

Lexical disambiguation is one of the most difficult problems in machine translation research.The authors discuss in details, together with many well-chosen examples, various means and techniques to help solve the problem, which have been developed and proved effective in the development of our English-Chinese JFY-Ⅳ MT system. They involve. 1. closed word group processing; 2. grammatical analysis; 3. semantic analysis; 4. context analysis; 5. analysis of referential words; 6. careful selection of more adaptable target equivalents; 7. recursive analysis of interdependent words, 8. giving two or more translations. With all these mothods complementing each other, most problems concerning lexical ambiguities will be adequately dealt with, leading to the goal of higy quality machine translation.