Introduction to the Special Issue on Machine Translation of Asian Languages

摘要

It has been two decades since the great paradigm shift in machine translation toward the use of statistical methods, that is, automatically learning translation rules between two languages from large collections of translated text. These two decades have seen a rapid evolution from the original word-based models proposed by IBM researchers to phrase-based models and now to methods that use formal grammars with linguistic annotation. The use of statistical methods, together with growing availability of data in the world's languages, has lowered the barrier of entry for researchers to work on new languages. While the majority of commercial machine translation systems have been built for European languages, two Asian languages have received special attention by the statistical machine translation community: Chinese and Arabic.