语义缺省对机器翻译质量的影响研究

Research on influence of semantk defaults on machine translation quality

语义缺省普遍存在于自然语料,缺省的内容则是世界知识。目前,机器无法推理出缺省的内容,导致机器翻译结果不尽如人意。为探究语义缺省对于机器翻译质量的影响,文章基于Jaszczolt的缺省语义学理论,以《中国日报》中的英文新闻标题为语料,其中以文译文为参考,对比加入世界知识前后的机器翻译质量,研究语义缺省对机器翻译的影响。研究发现,各类缺省均对机器翻译结果有影响,其中多义词的影响最大占比63%,各类缺省在加入世界知识后机译质量显著提高。其中,隐藏关系中对于语态、时态助动词的省略则不影响机翻结果。根据语料的缺省分类对语义缺省的框架做了进一步完善,同时总结了针对不同的类型缺省如何填补相应的世界知识,为提高机器翻译质量提供了新思路。

Semantic defaults generally exists in natural corpora,and the default content is world knowledge.At present,the machine cannot infer the default content,resulting in unsatisfactory machine translation results.In order to explore the impact of semantic defaults on the quality of machine translation,this paper,based on Jaszczolt's default semantics theory,takes the English news headlines in China Daily as the language material,in which the text translation is taken as a reference,compares the quality of machine translation before and after adding world knowledge,and studies the impact of semantic defaults on machine translation.It is found that all kinds of defaults have an impact on the results of machine translation,among which polysemy accounts for 63%,and the results of machine translation of all kinds of defaults have been significantly improved after adding world knowledge.Among them,the omission of voice and tense auxiliary verbs in hidden relations does not affect the results of MT.According to the default classification of the corpus,the framework of semantic defaults is further improved.At the same time,it summarizes how to fill the corresponding world knowledge for different types of default,which provides a new idea for improving the quality of machine translation.