面向多引擎融合技术的统计后编辑方法研究

摘要

This paper introduces the machine translation system is developed by our company, which integrates the SPE (statistical post editing) method to achieve multi-engine fusion technology. The purpose of this method is to solve the problem of adaptive translation, and to improve the quality of translation. It includes the text of the RBMT system generated by the source language is used to construct the Chinese parallel corpus, and then use the statistical method to train translation model and language model, and then extract the phrase table and N-gram text for the SPE translation engine decoding. This paper also presents a comprehensive performance evaluation, and finally provides a vision for the future work.%本文介绍了我单位研发的机器翻译系统融入统计后编辑方法后实现串行结构多引擎融合技术，该方法的目的是解决翻译领域适应性的问题，矫正规则为主导的系统翻译结果，提高翻译译文的质量。其中包括提取系统生成的译文作为源语言构建中文平行语料库，然后采用统计的方法训练翻译模型和语言模型，抽取短语表和N-gram文本用于统计后编辑翻译引擎的解码。文中给出了该方法综合性能评价，最后给出下一步工作设想。