The Design And Implementation Of Bidirectional Translation Memory Subsystem In Multilanguage Computer Aided Translation System

Although in recent years, machine translation technology development made alot of progress, but it is hard to live up to all of the quality requirements. Therefore,began to consider using machine translation, meanwhile, using computer-assistedtranslation technology-translation memory technology to improve the quality oftranslation. Translation memory technology to language complex and changeful, makefull use of various function of memory, new use of the existing translations, in manyrepetitive higher some areas using this technology, can avoid translator repetitivelabor, avoid a sentence appeared a variety of translation, keep the consistency of, cangreatly improve the efficiency translation.In this paper the research content translation memory technology around some ofthe key technology spread out:Sentence similarity calculation in translation memory technology is one of thecrucial technology. According to sentence similarity calculation, first to naturallanguage processing fields used in several similarity calculation method to therelevant research and analysis. And combined with Chinese to Uygur, Kazak, Kirgiz,and some of the Chinese characteristics and differences between, finally to adoptshortest edit distance between sentences to calculate the similarity.Memory is translation memory system based on the translation. The structure ofthe memory and management will directly affect the translation memory system at theend of the day the use effect. Therefore, before the design, the study of this part ofsome light on and analysis, finally designed a memory. Memory bank gathered in allthe sentences in XML language are as code language. Translation memory with"translation unit" form of storage, sentence by sentence and precise correspondence.The corresponding sentence by sentence marker on <tu>...<tu> of id to describe. The system by2010c#related modules realize the corresponding function. Memoryafter design, but also to memory bank for good management, including the memorybank to find the word, add sentences, delete the sentence, memory bank import,export and so on. On the outside, in this article uses the shortest edit distance sentencesimilarity calculation method, and on the basis of the method of generating the.The formation is a translation memory system to get high quality the key.Translation memory systems, to be translated sentence by sentence similarity iscalculated, according to the similarity of the structure as a model to construct thesentence pattern. When an input sentence, translation memory system through theinterior of the algorithm and the translation process, in the memory of a certainmatching. When the translation of sentences and the presence and degree of similarity(0%-100%), and the percentage of similarity is greater than or equal to me what wementioned above the threshold, the translator can be similar to the versions of thecorresponding improvement, delete, insert and other operations are to be translatedsentence translation.Finally, in translation memory technology developed on the base of themultilingual translation memory system. Through the similarity calculation ofsimilarity degree to get different processing to the corresponding translation work. Inorder to test the system developed in this paper, through a product version M630andM868version of the manual translation to investigate the use effect. The results showthat, in the repeat rate close to47%of the material, the translation of the increasedspeed for1day, that is the guarantee the quality at the same time, obviously improvethe translation speed.