Translingual Information Management by Natural Language Processing

Preface Translingual information management that can choose appropriate information and obtain useful knowledge from the ood of global information is being increasingly demanded. Among the related technologies, natural language processing is one of the most promising for meeting information needs because natural language processing can deal with documents that have essential roles in information. This research is aimed at developing a practical system for managing translingual information. The system is based on technologies for natural language processing. Translingual information management has a fundamental cycle of information needs and information search. In order to enhance the search quality and to accelerate the cycle, this research focuses on three core technologies: coordinate structure analysis, cross-language information retrieval, and information navigation. First, we propose a model for analyzing coordinate structure. This model provides top-down scope information on the correct syntactic structure by taking advantage of the symmetric patterns of parallelism. The analysis of coordinate structure creates a bottleneck when dealing with documents because most long sentences contain a coordinate structure. The model results in a high-quality search because it enables accurate analysis. Second, we propose the GDMAX method, which is a method for translating query terms for use in cross-language information retrieval (CLIR). CLIR is a key component in translin-gual information management. This method produces high-quality search results by choosing appropriate translation terms for CLIR. Finally, in order to make translingual information management ecient, we propose a method for information navigation that classies documents and enables a user to navigate by using 5W1H (who, when, where, what, why, how, and predicate) information. i ii Acknowledgments First and foremost, I would like to thank my supervisors, Prof. Hozumi Tanaka (TITECH, Japan) and Prof. Takenobu Tokunaga (TITECH, Japan) for their guidance, support and encouragement throughout the years of my Ph.D. studentship. I would also like to thank other members of my thesis committee, Prof. was conducted jointly with him, and without his advice and proposals I would not have been able to achieve the results reported here. I also extend thanks to Mr. Shin-ichiro Kamei for his invaluable advice and discussions from viewpoints of linguistic engineering. I am also grateful to Mr. Shin-ichi Doi and Mr. Kiyoshi Yamabana for their useful discussions and suggestions about CLIR and the DMAX method. I am also deeply appreciative of the valuable support of Mr. Eduard Hovy gave me global viewpoints on natural language iii iv processing and Kevin …