This article proposes an approach to the generation of guided tours over indexed information resources on user demand. It represents an alternative to more sophisticated hyperdocument generation techniques, in cases where very few is known about resource contents. The information space is modelled as a hypergraph of resource and resource descriptors. A spanning tree of descriptors is extracted from the line graph of this hypergraph to constitute the guided tour structure. We use a matching between spanning tree edges and a subset of transition resources to strengthen guided tour coherence and avoid redundancy. The model relies on an analysis of cognitive issues of guided tour consultation. KEYWORDS Hyperdocument generation, learning strategies, hypergraphs.