基于句对比较的自动获取翻译模板方法改进

Improvement of automatic acquisition of translation template based on sentence pairs comparison

翻译模板不仅是机器翻译的重要资源,而且是有用的语言学知识。使用词对齐结果改进了从实例中类比学习翻译模板的方法,并提出了带有函数关系的模板(TTFR)以获取语言之间的依赖关系,如"he…his"对。带函数关系的模板是一个更一般化的模板并能直接用于双向的机器翻译。实验结果表明改进的方法有效地克服类比方法的问题并且提高了自动获取模板的准确率。

Translation template is not only vital resource for machine translation,but also useful linguistic knowledge.Improves the analogy technique to induce translation template from given bilingual examples by filtering with the word alignment of the exam- pies,and presents Translation Template with Functional Relationship（TTFR） to capture dependencies such as ＂he...his＂ pairs and similar linguistic phenomena,which is a more general translation template and can be used in hi-direction machine translation directly.Experimental results show that the improved method breaks the limitation of the old learner and increases the accuracy of automatic inducing the translation templates.