关键词: 翻译模板，错误驱动，Nist评测

Abstract: 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 pap}r,hrecto-String method was applied to extract translation equivalents. Error-driven learning method was used to acctuire 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 0 o under 3-gram Nist assessment metric.

Key words: Translation templates, Error-driven, Nist assessment