Research And Implementation Of Chinese-Japanese Pattern-based Machine Translation

With the development of internet technogoly, communication between different cultures is becoming more and more frequent. As one of the most challengeable subjects in the field of Natural Language Processing, machine translation is paid much attention by the research institutes. This paper discusses the method of pattern-based machine translation between Chinese and Japanese based on the collaborative project between Northeastern University and FUJI XEROX Company. And we implement a real text translation system.Patten matching and phrase translation are two key factors which influence the performance of PBMT. The paper ameliorates the method of pattern-based machine translation through the direct matching between text and pattern, pattern-selected and pattern-nesting model. The performance of translation system is improved.The main content includes:(1)The paper ameliorates the performance of traditional PBMT by the means of direct matching between the given sentence and pattern as it avoids the difficult linguistic analysis for extracting pattern. In the translation model, form transformation will be done on the sentence to be translated according to candidate pattern. And then calculate the edit distance between sentence after form transformation and candidate pattern. Mapping relationship will be got using back-track algorithm.(2)The paper involves a pattern-selected model to avoid the problems that come from the miss-matching of patterns. Also if a sentence has more than one candidate pattern, the pattern-selected model will be helpful. The pattern-selected model considers edit distance between sentence after formal transformation and candidate pattern, the quality of phrase translation and pattern coverage as key factors effecting pattern selection.(3)The quality of the phrase translation influences the final translation result directly. A phrase translation system is set up to improve the quality of phrase translation.Pattern-based machine translation is used as a sub-translation engine of the phrase translation system, and this solve the problems caused by nesting patterns.The experiment shows us a satisfactory result.