| With the development and continuous expansion of internet information, we can get more resources from it. Experts have estimated that there are about3000languages. Consequently, in order to fully utilize the internet resources, machine translation has substantial sense.This paper mainly presents around following aspects:1) Word Alignment. Word alignment, based on the existence of sentence level alignment corpus, in order to discover word level relationships. It has important meaning for corpus based translation system, word disambiguation, translation dictionary.2) Phrase Alignment. Phrase alignment can reflects the alignment in phrase level between parallel corpus. Phrase alignment is an important part in translation knowledge acquisition.3) Syntax Structure Alignment. Structure alignment analyzes and summarizes about bilingual syntax structure, then extract alignment relationship. The syntax structure alignment result can be directly applied to phrase alignment.4) The Application of Translation Knowledge. After translation knowledge obtained from parallel corpus, they can be applied to machine translation. Finally, we can see the improvement of translation system.The mainly achievements in our research are as follows:1) In word alignment part, we proposed a multi-word alignment method based on N-Gram language model, which effectively resolves the one-to-multi and multi-to-one problems in word alignment. At last, we create a bilingual translation dictionary.2) In phrase alignment part, considering complication of the syntax structure, we present a phrase alignment by classification. We divide the phrase alignment process into Noun-phrase, Verb-phrase, Adj-phrase, and Adv-phrase. The experiment shows that our method can gets better result on precision than some traditional methods. We also implement a system can extract the translating phrase units.3) Lastly, we design a complete machine translation system based on the English-Chinese bilingual corpus. |
PDF Full Text Request