摘要 |
PURPOSE:To improve the accuracy of a machine translation system by selecting the optimum syntax structure by the experimental knowledge of a translator when plural kinds of syntax structure are obtained from the same text, and reflecting it on a knowledge base. CONSTITUTION:An inputted original text is stored in a buffer A17. Morpheml analysis is executed on the original text by a translation program 16, and a result is stored in a buffer B18. Thence. all kinds of syntax analysis trees configurable from the content of the buffer 18B are assembled, and they are stored in a buffer C19. The matching of those syntax analysis trees with a syntax rule are performed. At this time, a syntax priority rule exists in a memory 6, and the optimum one in plural syntax analysis trees can be selected. The syntax priority rule can be accumulated by a learning system. and when the plural syntax analysis trees are generated by applying machine translation to one original text and plural translation texts are outputted based on them, priority is attached on the syntax analysis tree included in the optimum translation text selected by the translator. |