摘要： 根据维吾尔语形态变化丰富的特殊性，搭建一个基于Factored的维汉机器翻译系统，将Factored系统和基于层次短语的Joshua翻译系统以及Moses中基于句法的翻译模型进行系统融合，构建混淆网络。提出一种词级和句子级联合融合的维汉机器翻译方法，利用一致性网络进行词级融合，并采用最小贝叶斯算法进行句子级融合。实验结果表明，联合式多引擎方法能提高1.72%个BLUE-SBP值。
关键词:
机器翻译,
Factored短语,
多引擎,
维汉,
系统融合
Abstract: According to the agglutination of Uyghur, a factored-based Uyghur-Chinese translation system is constructed. A confusion network is explored which combines factored-based system, hierarchical phrase-based system Joshua and syntax model in Moses. A new translation method in Uyghur-Chinese translation combining word level and sentence level is proposed. It makes use of MBR to merge sentence level results. Experimental results show that the measure can improve BLUE-SBP by 1.72%.
Key words:
machine translation,
Factored phrase,
multiple engines,
Uyghur-Chinese,
system fusion
中图分类号:
宿建军, 张小燕, 吐尔洪?吾司曼, 李晓. 联合式多引擎维汉机器翻译系统[J]. 计算机工程, 2011, 37(16): 179-181.
QI Jian-Jun, ZHANG Xiao-Yan, TU Er-Hong-?Wu-Ci-Man, LI Xiao. Joint Multiple Engines Uyghur-Chinese Machine Translation System[J]. Computer Engineering, 2011, 37(16): 179-181.