This paper is about an optimization method to detect 'semantic exceptions' in context aware systems, which means the kind of exceptions considering application semantics. To provide user-friendliness and expressiveness for program developers, in addition to efficiency in detection, intuitive descriptions like 'too near(),' and 'too fast()' are supported, and transformed into an intermediate form called ILSE (Intermediate Language for Semantic Exceptions.) We focus on the optimization process, specifically when the opt-out parameter values in too near() are filled with system-provided values in the ILSE form. This paper shows efficient algorithms to find out optimal values for sensing rates, some of which also keep decent promptness in response to the exception. © 2013 IEEE.