【摘要】：在統(tǒng)計機器翻譯中融入有價值的句法層面的語言學(xué)知識,對于推動統(tǒng)計機器翻譯的發(fā)展具有重要的理論意義和應(yīng)用價值。提出了三種由簡到繁的將雙語最大名詞短語融入到統(tǒng)計翻譯模型的策略,整體翻譯性能逐步上升。Method-Ⅲ采用分而治之的策略,以硬約束的方式在統(tǒng)計機器翻譯中融入最大名詞短語,并在雙語最大名詞短語層面上融合了短語翻譯模型和層次短語模型,對翻譯系統(tǒng)的改善最顯著。所述策略顯著提高了短語翻譯模型的質(zhì)量,在復(fù)雜長句翻譯中,Method-Ⅲ的BLEU值比基于短語的基線翻譯模型提高了3.03%。
[Abstract]:The integration of valuable linguistic knowledge at the syntactic level in statistical machine translation is of great theoretical significance and practical value in promoting the development of statistical machine translation. In this paper, three strategies from simplicity to complexity are proposed to integrate the bilingual maximum noun phrases into the statistical translation model, and the overall translation performance is gradually improved. Method- 鈪，

本文編號：2408540