基于混淆网络的机器翻译系统融合研究-模式识别与智能系统专业论文.docxVIP

摘要率。实验证明改进词对齐质量的系统融合可以比经典的词级别系统融合提高 摘要 率。实验证明改进词对齐质量的系统融合可以比经典的词级别系统融合提高 0．1．O．5个BLEU。 3、改进混淆网络解码方法的系统融合研究 改进混淆网络解码方法对多个混淆网络进行解码，是针对目前的混淆网络解 码方法过分依赖参考句子、调序能力有限的改进。本文提出以重评分和最小贝叶 斯风险解码两种方法对多个混淆网络进行一致性解码。实验证明改进混淆网络解 码方法的系统融合可以比没有改进的融合方法提高0．5个左右的BLEU。 关键词：机器翻译、系统融合、词对齐、最小贝叶斯风险、混淆网络、重评分 Ⅱ AbstractABSTRACT Abstract ABSTRACT Recently，system combination has become one of the research hotspot in machine translation．System combination is a kind of technology to make use of different kinds of machine translation systems to improve translation quality．The present system combination methods Can be categorized into two major types： sentence level system combination and word level system combination．Sentence level system combination is based on minimum bayes·risk decoding and word level system combination iS achieyed in form of confusion networks．Owing to the advantage of improving translation quality stably and remarkably,word level system combination has become the popular method in system combination． Sentence level system combination based on minimum bayes-risk decoding needs to define the loss function which Can affect the combination performance．And word level system combination first needs to align many different machine translation system’S outputs and then builds a confusion network based on outputs’word alignment．At last the best combination result is extracted from the confusion network． Word alignment quality and confusion network decoding are vital to word level system combination’s perfofinance． The current sentence and word level system combination methods are analyzed and summarized comprehensively in this thesis．Three different kinds of loss functions based on machine translation evaluation are defined in sentence level system combination to test their effect on combination resua．Considering the word alignment of current word level system combination methods do not take language information into account，new methods which aim to improve the word alignment quality