首页>
外文期刊>Kwartalnik Elektroniki i Telekomunikacji
>Analiza mozliwosci i ograniczen systemow translacji automatycznej wspomaganej przez czlowieka na przykladzie systemu tlumaczacego z jezyka szwedzkiego na polski
【24h】
Analiza mozliwosci i ograniczen systemow translacji automatycznej wspomaganej przez czlowieka na przykladzie systemu tlumaczacego z jezyka szwedzkiego na polski
Machine translation is a science that delivers the knowledge how to program the computers, so as they were able to translate between human languages, for example, between Danish and Bulgarian. It may be amazing, but the field of machine translation is almost as old as the invention of computer itself. In 1949 an American scientist Warren Weaver sent the memorandum to The Rockefeller Foundation (American institution supporting the scientific research), in which he demanded starting the research on the automation of translation between natural languages. Warren Weaver was inspired by cpyptographic techniques, which were developed very strongly during the years of The Second World War, and he thought that there existed some fundamental similarities between these cryptographic techniques and the process of translation between human languages. But it soon appeared that the automation of translatin is a far more difficult and far more interesting from the scientific point of view than had ever been thought before. After 50 year of intensive scientific research machine translation for a general language is still far away from its satisfactory solution. The high-quality machine translation system for unrestricted text has always been an unachievable goal for the computer scientist working in the field of automatic translation between human languages. And maybe, because of the reasons of fundamental nature (the lack of possibility of constructing an algorithm equivalent to the creativeness of the human mind) human translators will never be eliminated by computers totally, and high-quality machine translation for unrestricted text will forever remain the Holy Grail of scientific research. But, by using various machine translation techniques one can try of course to approach as close as possible to this unattainable goal. Quite recently the example-based machine translation technique has emerged as a very serious and tempting alternative to the existing system that are mainly based on the knowledge developed in the field of computational linguistics. In the paper the implementation of example-based machine translation technique in the system, which translates from Swedish into Polish is proposed. In order to use the example-based machine translation technique for the Polish language, which posses very specific prammatical features, so different from the West-European languages, this author proposed a thorough modification of this technique that allows to take into account the flexion nature of the Polish language. The results obtainde so far are very promising and show that the usage of example based machine translation technique for the Polish and Swedish language pair is a step made in the right direction. But, one cannot forget that the final success depends strongly on the dimension of the database of translation examples. The effective constructing of such database requires a lot of work and time. In fact, it is a task for a quite big teem of trained linguists and computer scientist, who basing only on the great bilingual corpus would be able to extract all the necessary and most frequently used translation examples.
展开▼
