Annotations are an important way in which social machines interact with data. Most systems that have been developed for data annotation assume a two-level structure in which annotations are superimposed on, and separate from, the data itself. However, there are many cases in which annotations are themselves annotated, such as when discussion threads in emails and newsgroups make it possible to attach comments to other people's comments.
The Data Annotation project explores the challenges associated with annotated databases: We have developed hierarchical models of annotation in which there is no absolute distinction between annotation and data, and explored how annotations can propagate through queries in such a model. Annotated databases pose new challenges in query optimisation, as propagation of annotations demands that we reconsider many fundamental tasks, such as query containment.
