基于错误驱动的翻译模板自动获取  

Automatic Acquisition of Translation Templates Based on Error-driven

作　　者：张春祥[1] 梁颖红[2] 于林森[3] 

机构地区：[1]哈尔滨理工大学软件学院,哈尔滨150080 [2]苏州市职业大学计算机系,苏州215104 [3]哈尔滨理工大学计算机科学与技术学院,哈尔滨150080

出　　处：《计算机科学》2010年第7期229-232,共4页Computer Science

基　　金：国家自然科学基金(60903082);黑龙江省教育厅科学技术研究项目基金(11541045);哈尔滨理工大学青年科学研究基金(2008XQJZ017);江苏省现代企业信息化应用支撑软件工程技术研究开发中心项目(SX200907)资助

摘　　要：翻译模板自动获取是提高MT系统译文输出质量和领域快速移植能力的关键性因素。利用Tree-to-String方法来抽取等价对,使用错误驱动的学习方法来获取翻译模板。将获取的模板用于MTS2005中,同时对其译文质量进行开放测试。实验结果表明:所提出的模板获取方法的性能要好于传统方法,当新模板加入系统原模板库后,开放测试语料的3元Nist评测分数提高了3.41%。Automatic acquisition of translation templates is important for MT system to improve its translation quality and its ability of adapting to new domain. In this paper,Tree-to-String method was applied to extract translation equivalents. Error-driven learning method was used to acquire templates, These templates were applied to MTS2005, and open test experiments were conducted for its translation quality. The experiment results show that the performance of new method in this paper is better than that of old method. Combination of new acquired templates and original ones makes assessment score of open test corpus improved by 3. 41% under 3-gram Nist assessment metric.

关 键 词：翻译模板 错误驱动 Nist评测 

分 类 号：TP391.2[自动化与计算机技术—计算机应用技术]