The CAS 11th Five-year Informationization Construction Program will build a huge data application environment, which is composed of 51 institutions. It is necessary to build a cross-search tool to help people get useful information from those massive data. The traditional cross-search technology doesn't support the professional type of scientific data, which brings difficulty in data formatting and data presentation. This paper proposes a cross-search framework which supports these professional types. This framework uses data modeling middleware to publish scientific data and metadata on the basis of the universal data exchange format of the profession type, and uses template technology to display these data flexibly. The framework also manages professional type as a plug-in, which has good extensibility. This paper also introduces the application of the cross-search system Voovle, based on this framework.