Tags and Dependencies: an Integrated View of Document Annotation

Metadata provided by annotations are an essential prerequisite for the realization of the semantic web. The manifold of metadata uses on the other hand also implies an abundance of different annotation types and formats, each requiring a different semantic treatment of the data. For semantically rich documents this results in a hybrid mixture of metadata about one and the same document. Further metadata diversity arises, if more than one author contributes to the annotation as it is typical for the now popular social tagging systems. Documents however, despite this heterogeneity show common structural properties that can be classified as logical, conceptual, and referential. There are intrinsic dependencies within and across those structures that have to be made explicit. We argue that the dependency structure as an explicit annotation is essential for any semantically rich document in order to be better understandable not only for man but also for machine.