人机翻译英语被动句相似度的向量距离测算

摘要

基于语料库翻译学和文本挖掘的方法，提出了一种将向量空间模型与主成分分析相结合的向量距离算法，用以测算在线机器翻译（“在线机译”）与人工翻译在对应翻译英语被动句上的相似度。测算结果表明，人工译本和在线机译都体现了英语被动句翻译的不对等性，均将少量英语被动句译成汉语被动句。研究结果或为译本对比及文本相似度的测量提供一种新的数理统计算法，从而更加客观、科学地对比译本之间的异同。%By combining corpus -based translation studies with textual data mining methods,this research proposes a new vector distance algorithm by combining Vector Space Model (VSM)and Principal Component Analysis (PCA),to measure the similarity between online translation and human translation of English passives into Chinese passives in a formally parallel manner.The results show that not many English passives are translated correspondingly into Chinese passives.The four human translated texts studied are close to each other but far from online translation in vector dis-tance,indicating a low similarity between human and online translations.Conclusively,the results and method of this study can be applied to translation comparison and provide a new statistical algorithm for measuring textual similarity so as to study and compare different translated texts more objectively.