Current desktop-GIS software cannot answer users' spatial questions directly. The GIS functionality is hard to identify and use without specific training of GIS skills because of the complex hierarchical organization and the gap between users' spatial thinking and systems' implement descriptions. In order to bridge this gap, we propose a semantic framework for designing a question-based user interface that integrates different levels of ontologies (spatial concept ontology, domain ontology and task ontology) to guide the process of extracting the core spatial concepts and translating them into a set of equivalent computational or operational GIS tasks. We also list some typical spatial questions that might be posed for spatial analysis and computation. The principle introduced in this paper could be applied not only to desktop-GIS software but also to web map services. The semantic framework would be useful to enhance the ability of spatial reasoning in web search engines (e.g. Google semantic search) and answering questions in location-based services as well (e.g. iPhone Siri assistant).