基于对等模式的汉-英译文调序  

Reordering the translations from Chinese into English based on peer patterns

作　　者：张春祥[1] 赵铁军[2] 卢志茂 高雪瑶[4] 

机构地区：[1]哈尔滨理工大学软件学院,哈尔滨150080 [2]哈尔滨工业大学计算机科学与技术学院语言语音教育部-微软重点实验室,哈尔滨150001 [3]哈尔滨工程大学信息与通信工程学院,哈尔滨150001 [4]哈尔滨理工大学计算机科学与技术学院,哈尔滨150080

出　　处：《高技术通讯》2013年第1期29-34,共6页Chinese High Technology Letters

基　　金：863计划（2006AA010108）和国家自然科学基金（60903082,60975042）资助项目.

摘　　要：为了提高机器翻译质量，提出了一种基于对等模式的汉-英译文调序方法：从短语翻译对中抽取汉．英语序对应关系，利用语言学特征和错误驱动学习相结合的方式获取对等模式，使用对等模式来改变汉语句法树结构，使其生成的译文符合英语语序要求。使用该方法对500个汉．英双语句对中的汉语句子进行调序的实验结果表明，词链交叉率降低了10．56％。经过调序之后，汉语句子的译文质量有所提高。Translation reordering in machine translation was studied for improving the quality of translation output. A method for reordering the translations from Chinese into English based on peer patterns was put forward below： the correspondent relationships between Chinese word orders and English word orders were extracted from phrase trans- lation pairs learning, w word orders pairs in an tion quality , and then, the peer patterns, obtained based on the combination of linguistics features and error-driven ere used to change the structures of Chinese parsing trees, making the translations accord with English . The new method was used to reorder Chinese sentences of 500 Chinese-English bilingual sentence experiment, and the results showed the crossing rate of word links degraded 10.56%, and the transla- was increased because the Chinese sentences were reordered.

关 键 词：对等模式 译文调序 短语翻译对 词链交叉率 译文质量 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]