Accessing Document Data Sources Using Referring Expression Types.

We show how JSON documents can be abstracted as concept descriptions in an appropriate description logic. This representation allows the use of additional background knowledge in the form of a TBox and an assignment of referring expression types (RETs) to certain primitive concepts to detect situations in which subdocuments, perhaps multiple subdocuments located in various parts of the original documents, capture information about a particular conceptual entity. Detecting such situations allows for normalizing the JSON document into several separate documents that capture all information about such conceptual entities in separate documents. This transformation preserves all the original information present in the input documents. The RET assignment contributes a set of possible concept descriptions that enable more refined and normalized capture of documents, and to more crafted answers to queries that adhere to user expectations expressed as RETs. We also show how RETs allow checking for a document admissibility condition ensuring that each document describes a single conceptual entity.