Introduction to statistical machine translation
暂无分享,去创建一个
The automatic machine translation of a text in one natural language into another natural language is a task which been considered for at least as long as computers have existed. A wide variety of techniques has been considered for this task, although none has proved outstandingly successful, at least judging by the performance of current translating systems. It seems worthwhile, therefore to consider the relevance of techniques not seriously considered before, just in case some additional insight can be obtained into the translation process. The approach of treating language translation as a task in decoding, familiar in many applications of Information Theory, is briefly described here. This approach has been found very successful in Machine Speech Recognition, and it may be that some of the techniques used there can also carry over to the Machine Translation domain. Some preliminary results suggest that the approach may be capable one day of performing certain translation tasks, although it is by no means a commercially practical proposition at the moment.