Study and implementation on key techniques for an example based machine translation system

We analyze several key parts of machine translation, study the framework of machine translation systems and dig out the factors that should be considered during machine translation modeling. We dissect the sentence segmentation, sentence alignment, translation knowledge acquisition, translation fragment selection and the role of translation generation in the machine translation process in detail. After comparing some common sentence alignment model, we propose a comprehensive word-based sentence aligned model. We also put forward a basic translation fragment method, which can automatically get basic translation fragment library from bilingual corpus using the result of vocabulary alignment. The method can extract translation fragments in any bilingual corpus since the knowledge of sentence structure analysis is not used, resulting in advantage of language independence. Finally we design and implement an example-based machine translation system. Experimental data show that the example-based machine translation system proposed can carry out translation well, and has translation knowledge automatic acquisition and language independence features.