The emergence of digital documents renews the relation between annotation and indexing, thanks to the processing it enables. We propose a unified framework (so called index) in which these two operations complement each other and support the interplay of bidirectionnal moves between documents and a semantic representation. This framework shows a good theoretical coverage and accounts for different activities. Its practical efficiency is illustrated on an ongoing experiment, which aims at indexing texts describing policies, which issue is explained and for which we present an original index structure. This example illustrates how that structure can be turned into a navigating space connecting a text and a semantic model, and how, on this ground, new functionalities can be built which enlarge the uses of digital documents.