Japanese Long Sentence Parsing

Japanese syntactic analysis is the kernel content in natural language processing and machine translation. This dissertation is based on Japanese-Chinese Machine Translation System and discusses the Japanese syntactic analysis method. In this paper, I focus one of the difficult problems in Japanese paring-long sentences syntactic analysis.The sequence of Japanese is very freedom. In practical text, the number of long sentences is in great proportion. The research results have indicated that it is difficulty in analyzing the long sentences only resorts to the method used in short sentences. The analysis of sentence that beyond 50 words is very difficult. When the number is above 80 words, the parse is always unsuccessful. So in long time, the paring of long sentences is the difficult problem of Japanese understanding and translation.The cause in forming of the long sentences can be generalized into the present of coordinate component and multi related sentences. Focused on the determination of the coordinate relations, Sadao Kurohashi proposes the algorithm of path analysis to settle the part of problems. But this algorithm often emerges with the wrong analysis problem in treating with the asymmetric configuration. After that SATOSHI SHIRAI proposes the method of subordinate clauses analysis. This method increases the syntactic analysis precision of long sentences with multi related sentences.This paper analyzes the algorithm proposed by Sadao Kurohashi. Focused on the demerits of the algorithm I discuss the improved way combined the subordinate clauses analysis. At the end I integrate the merits of the two methods and propose the practical resolution.This method has been applied to "NIHAO" Japanese-Chinese Machine Translation System to authenticate the validity. The result is that the parsing precision of long sentences is improved greatly and the method has achieved the level of practical application.