【作者】 张学； 黄德根；

【Author】 Zhang Xue Huang Degen Department of Computer Science and Technology, Dalian University Of Technology, Dalian 116024

【机构】 大连理工大学计算机科学与技术系；

【摘要】 在EBMT(Example-BasedMachineTranslation)系统中将翻译实例泛化为翻译模板,可以有效的减少实例的存储空间,提高实例的检索效率,而实例匹配更是直接关系到了EBMT系统的翻译质量。本文提出了一种利用汉语句子的表层句法信息和词汇语义信息,从实例中提取多级翻译模板的方法。多级翻译模板可以在模板匹配中关联计算相似度并独立进行翻译。该方法产生的翻译模板规模小,可以有效的降低相似度计算的复杂度并提高精确度。更多还原

【Abstract】 Example generalization in the EBMT (Example-Based Machine Translation) system is very meaningful for both reducing the storage space and improving the searching efficiency of examples, and example matching directly affects the quality of the EBMT system. In this paper, an approach which utilizes the information of syntactic structure and lexical meaning of the Chinese sentences to extract the multilayer translation template from example is applied. With the multilayer translation template, all layers would be computed together in the similarity computation and each layer would be translated individually. The translation template extracted by this approach is small-scaled, it can effectively reduce the complexity of the similarity computation and increase the precision of the example marching.更多还原