The Research And Implementation Of An English-Chinese Machine Translation System

With the growth of international communication and development of the World Wide Web, the language barrier is becoming a serious problem. People give more and more expectation on Machine Translation (MT) as a technology for overcoming the language barrier. However, Due to the inherent complexity of the natural language, machine translation is still a great challenge until now. Machine Translation has been explored for nearly fifty years. Researchers have found that it is impossible to use only one method to process the various complex cases. Current MTS integrates multiple method and strategy. At least, in differ phrase uses differ method and strategy. This paper proposes a model of rule-based English-Chinese Machine Translation System. The MTS consists of translation engine and machine dictionary. On the control-strategy of the translation, the system uses the multi-phase analysis methods. According to various kinds of ambiguity in the language, the machine translation procedure is divided into six phases: preprocessing, morphology analysis, POS tagging, phrase analysis, clause analysis and translation choice. Every phase uses different method and strategy. The translation process doesn't create complete the grammar tree, but creates a list of words and a list of phrases. Clause is basic transfer unit. A sentence is viewed as recursion of clause. A clause is recognized through matching patterns of sentence and is translated into Chinese. On the system of knowledge of the MTS, the common and procedural knowledge of the language is embedded in the translation program, but the individual, descriptive, restraining knowledge of the language is stored in the machine dictionary. Every word entry, except of POS attributes, subcategory attributes, semantic attributes and collocation, includes rules and conditions of disambiguity. The methods of knowledge organization can avoid the conflict from the different knowledge granularity. This system is developed and debugged on Visual C++ platform under Windows environment. The result of preliminary experience is satisfying. If the machine dictionary is large enough, this English-Chinese can work well, especially in a limited field.