ドメイン知識に基づく UNIX 日本語インタフェースの構築

概要

論文の詳細を見る
One of the current research for human-computer interface is natural language interface. It reduces users' amount of study of computer systems and helps users with little knowledge of computer to use computer systems. This paper describes the construction of a Japanese user interface for UNIX file handling operations. The user interface accepts Japanese sntence and generate a sequence of UNIX commands. This system includes specific knowledge of several domains : knowledge of functions of UNIX commands, methods for command combinations, and functions of Japanese words on a specific domain. Domain specific knowledge is used in two part of the system, the syntax and semantics analizer and the code generator. The syntax and semantics analizer selects an appropriate description for each word from the dictionary based on word restrictions and generates internal representations for the input Japanese sentences. The code generator transforms an internal representation into a UNIX command sequence, and when needed, uses pipe and redirection to modify the commands. During code generation, the system checks the commands execution results based on command dictionary description.
山口大学の論文