Proc. Berliner Infromatik-Tage, 72-81, 1989This paper describes a general purpose text database management system SIGMA that exploits very fast and efficient one-way sequential processing of text files. For the implementation, a fast pattern matching algorithm which can deal with multiple keys is designed to realize commands for searching, replacing and extracting keys. The algorithm also copes with Japanese texts containing both 1-byte and 2-byte characters with shift codes. A new file system is also designed to manage large text files efficiently. The system SIGMA is now working at the Computer Center of Kyushu University for large scale text detabases, natural language analysis and bibliographic retrievals