Alignment Analysis of English Chinese Bilingual Corpora based on CRFS Model

Abstract

Bilingual corpus alignment analysis, as an important research method of machine translation, is of great value as a source of translation knowledge. In view of this, this study carries out the research on English Chinese bilingual corpus alignment based on CRFs model. That is, on the basis of chunk aligned corpus, combined with the respective language characteristics of English and Chinese, the word alignment between chunks is realized with the help of CRFs model. The results show that the accuracy and recall of maximum entropy are 45.77% and 45.58% respectively, the accuracy and recall of log linear model are 45.73% and 45.68% respectively, while the accuracy and recall of CRFs model are 47.99% and 47.94%. It indicates that CRFs model can effectively alleviate the asymmetric problem of English and Chinese database alignment, also the alignment error rate of CRFs model is reduced accordingly. Therefore, it has a good effect on the alignment of English and Chinese bilingual corpora.