一种面向汉英口语翻译的.pdfVIP

一种面向汉英口语翻译的 双语语块处理方法 程葳 赵军 徐波 刘非凡 （中国科学院自动化研究所模式识别国家重点实验室 北京 100080） 摘要：基于语块的处理方法是近年来自然语言处理领域兴起的一条新思路。但是，要将 其应用于口语翻译当中，还需按照口语特点对涉及双语的语块概念做出合理界定。本文在已 有单语语块定义的基础上，根据中、英文差异和口语翻译特性，从句法和语义两个层次提出 了一种汉英双语语块概念，并对其特点进行了分析。同时，针对中、英文并行语料库，建立 了一套计算机自动划分与人工校对相结合的双语语块加工方法。应用该方法，对汉英句子级 对齐的口语语料进行双语语块划分和对整，并以此为基础进行了基于双语语块的口语统计机 器翻译实验。结果表明，本文提出的双语语块定义符合口语翻译的实际需要，使用基于双语 语块的语料处理方法，能有效地提高口语系统的翻译性能。 关键词：统计机器翻译；口语翻译；语料库；语块 中图分类号：TP391 Bilingual Chunking for Chinese-English Spoken-language Translation CHENG Wei, ZHAO Jun, XU Bo and LIU Fei-Fan (National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences Beijing 100080) Abstract ：Chunking is a useful step for natural language processing. The paper puts forward a definition of bilingual chunks for Chinese-English spoken-language translation, based on both the characteristics of spoken-language and the differences between Chinese and English. Some special features of these chunks are also analyzed. Based on the definition and analysis, a method is proposed to segment the chunks in bilingual corpora. This method includes two steps of the automatic chunking and the manually modification. Using this method we got a chunk-aligned Chinese-English bilingual corpus. A series of chunk-based statistical machine translation experiments are then conducted which shows that the proposed definition and the bi-chunking method can lead to great improvement to the quality of the Chinese-English spoken-language translation. Keywords ：statistical machine translation; spoken-language translation; corpora; chunk 本文受国家“973”项目G1998030501A-06、国家自然科学基金重点项国家自然科学基金项 助；作者：程葳，博士研究生，主要研究方向为口语翻译；赵军，博士，主要研究方向为自 然语言