A mediator system is a kind of a meta-search engine that provides a seamlessly integrated search service for diverse search engines (collections). Since collections of a mediator system are geographically distributed, its performance is mainly influenced by the data transmission time between the mediator and its collections. Existing mediator systems employ a result cache that is composed of the results of previously issued queries to reduce this transmission time. However, these systems do not support a general Boolean query model but only simple ones such as a conjunctive query, a single keyword query, and or so. In this paper, we propose a new method to efficiently process general Boolean queries using a result cache for mediator systems. Also presented is a way to semantically partition the given result cache in order to reduce the complexity of inference.