日中机器翻译中的可保留歧义关系

On reservable structural ambiguities in Japanese-Chinese machine translation

着重阐述了日中机器翻译中可保留歧义关系（ＲＳＡ）这一新概念及其处 理手法．所谓可保留歧义关系是指即使不消除原输入文中授受构造上的歧义 性也能生成反映原文意思的译文关系．具体而言，就是研究由并列助词［］、 连体助词［の］和名词构成的名词短语及由用言连体形构成的句子中授受构 造上的可保留歧义性；并在此基础上提出了一个新的翻译方法．从技术论文 的２ ９１９个句子中抽出了含有上述可保留歧义关系类型的句子８０６个进行了 翻译实验，正确率为９７．７％．从而确认了所提方法的有效性．

A new concept, Reservable Structural Ambiguities (RSA) in Japanese-Chinese machine translation,is presented. The RSA are structural ambiguities of Japanese which can be translated into Chinese without being resolved. Usually ,when translating one language into the other which doesn't belong to the same 'language family', structural ambiguites must be resolved. However, Japanese and Chinese don't belong to the same family,their differences being mostly on sentence structure. However ,some parts of their sentences are similar ,making RSA possible ,and this paper will focus on them. Some RSA patterns will be discussed and a method for generation of Chinese from Japanese sentences with RSA will be given. An experimental system based on this method has been constructed and an experiment on 806 sentences which have RSA, taken from 2 919 sentences from technical articles, has been carried out. The result is a correct translation rate of about 97. 6%, showing that the proposed method is quite effective.