Corpus-Based Machine Translation Of WebPages--A Suggestion On MT Model And Strategy To Disambiguation

Machine Translation has a history of almost 60 years since the advent of computer. Its development has experienced many stages, and got great achievements. With the development of Information Technology, Internet has become a new mass media. However, the fact that most of the Internet resources are presented in English has been a great barrier for us Chinese users to obtain updated information in time. Although many WebPages machine translation (MT) software have been developed all over the world, its quality is not so good as expected.One important phenomenon should not and should never be neglected in MT system, that is, ambiguity. For tagged text, with the 4-step method, ambiguity can be greatly reduced. This thesis focuses on the strategy to disambiguation on pure text. The author first describes different kinds of ambiguity occurring in translation, and then proposes a corpus-based strategy to this problem. Only when the two problems are solved satisfactorily can the quality of MT be greatly improved.