This research investigates spatial information retrieval in architecture, based on an efficient metadata that is crucial for content-based retrieval. SpaceCore is the metadata schema for spatial information on a floor plan implemented with RDF (Resource Description Framework). To generate metadata, we analyse the factors of spatial information, and then construct a data model containing the content and structure of spatial information. To exploit the metadata, we suggest various query operations with possible predicates and interfaces. SpaceScope, the spatial information retrieval system, consists of three modules: an input data module, a query module, and a browse module. As to implementation, we have already developed the real time plan-structuring system, called “Vitruvius”,  as an inputting and modelling data module (using C++). Based on the database, the system performs the similarity analysis and then evaluates the alternative plans with the retrieval goals. At the end of the paper we discuss the potential of SpaceScope in the schematic design stage of the architectural design process, possibly improving and triggering the designeris conceptual approach against a logical background and facilitate rapid decision-making after full consideration.