Untranslatability And Irreversibility In Machine Translation

Nowadays, with the speeding up of globalisation, both international communication on business and trade, and international academic conferences are increasing day by day. More and more documents and other materials need to be translated from one language into another. Under this circumstance, traditional manual translation appears insufficient, so people turn to machine translation. Machine translation, shorted as MT, which is the application of computers to the translation of text from one natural language into another. Although at the very beginning, people's erosion to MT is optimistic, active and positive, and a lot of funds was invested to it, there were hundreds of MT research institutes around the world, owing to the characteristics of MT, namely, complex, interdisciplinary and abstruse, semantic barricade brought forth in ALPAC (automatic language proceeding advisory committee) has not been solved up to now, high quality and automatic text is still a dream that all of the researchers are longing for but they never attained. Besides the underdeveloped technology, unreasonable assignment is another important factor. Traditional MT researchers only divided the to-be-translated texts into two categories: literary text and non-literary text. They think that the non-literary text can get ideal translated-text through machine while the literary text cannot. However they neglect that there are always literary elements in non-literary text.So the author of this thesis brings forth a suggestion on how to improve the quality of machine-translated text under the circumstance of present technologies. The author firstly analyses the to-be-translated text according to some linguistic rules, then the advantages and disadvantages of both brain and machine are explored. According to the characteristics of brain and machine, the author divides the to-be-translated text from a new perspective. Combining untranslatability and irreversibility with the characteristics of machine, all the to-be-translated texts can be divided into three categories: translatable and reversible, translatable but irreversible, untranslatable and irreversible. As far as the first category is concerned, machine can translate it very well under the condition that the sentence structures are not complex. But as to the second category, machine can only translated part of it effectively, for machine cannot translate the untranslatable elements in a text. For the last category, machine can do nothing of it. Thus, in order to improve the readability of machine-translated text and improve the efficiency of machine, the author suggests choosing the first category for machine to translate and assigning what machine cannot translate or cannot translate well to manual translation.