半结构化的基于实例的机器翻译semi-structured example based .pdfVIP

半结构化的基于实例的机器翻译 王海峰 刘占一 吴华 东芝(中国)研究开发中心 北京 100738 E-mail: {wanghaifeng, liuzhanyi, wuhua}@ 摘 要：本文介绍了一种半结构化的基于实例的机器翻译方法。该方法将双语翻译实例表示为半结构化的树 串映射，一个树串映射由三部分组成：源语言分析树、目标语词串、及它们之间互译词的对应关系。基于这 种树串映射进行机器翻译时，首先使用源语言分析器得到待翻译句子的分析树；然后在翻译实例库中搜索与 该分析树最匹配的树串映射森林；最后，使用线性对数生成模型产生译文。使用 BLEU 作为机器翻译评价准 则的实验结果表明，本文方法显著优于著名的基于短语的统计翻译系统Pharaoh 。 关键词：树串映射，翻译实例，基于实例的翻译，机器翻译 Semi-Structured Example Based Machine Translation WANG Haifeng, LIU Zhanyi, and WU Hua Toshiba (China) Research and Development Center, Beijing, 100738 E-mail: {wanghaifeng, liuzhanyi, wuhua}@ Abstract: This paper describes a semi-structured example based machine translation (EBMT) method. In this method, the translation example is represented as a Tree String Correspondence (TSC), which is a triple consisting of a parse tree in the source language, a string in the target language, and the correspondence between the leaf node of the source-language tree and the substring of the target language string. An input sentence is first parsed into a tree. Then the TSC forest which best matches the input tree is searched for. Finally the translation is generated using a log-linear generation model to combine the target-language strings of the TSCs. Using BLEU as a metric, our method is significantly better than Pharaoh, which is a famous phrase-based statistical machine translation system. Keyword: tress string correspondence, translation example, example based machine translation, machine translation 1 前言 基于实例的机器翻译（Example Based Machine Translation ，简称 EBMT ）的本质是“以 翻译实例为基础，基于相似原理的机器翻译”，其利用的主要知识源是预处理过的双语语料 和翻译词典[1] 。典型的 EBMT 模型通常包括三个模块：在翻译实例库中搜索匹配片段；确 [2] 定相应的译文片段；重新组合译文片段以得到最终翻译 。 本文介绍了一种