As requirements on switching systems become complicated, it is essential to understand user's request correctly. A model of user request understanding is proposed in this thesis and a program is developed by which user's requests in natural language can be analyzed and some executable services are found from the view point of switching system operator.
To simplify the analysis procedure, user's request is mapped into one of four levels according to the clearness of the request. The simplest request is asking a connection by giving a phone number. A less simple request is asking by giving unconfusable index such as name or identification number. A complicated request which may need historical data for the user is stating direct goal such as reservation of flight. The system then consults the user's historical data and select a specific place corresponding to the direct goal. The most complicated one which may need intelligent reasoning is stating the user's goal in an indirect way. For example, the user states just that he has to travel, then the system reasons out he may need to call a travel agent. A request of a higher level is converted into a lower level until it becomes the simplest one.
Schank's McELI is adopted to analyze user's request. To represent the typical frame of communication service, connection-frame is added to existing conceptual dependency structure. Although the program cannot achieve a satisfiable result because of its insufficient collection of knowledge, it demonstrates the feasibility of such a system.