The structure of abstract document objects

Underlying every document processing system is a model of the document. For many applications a simple model, such as a long string of characters, is adequate. However, more expressive models are desirable for more demanding applications that involve complex textual material and also nontextual objects, such as mathematical notation, tables, and figures. In this paper we present an expressive model, suitable for paper and electronic documents, that is based on a graph-like structure. The principal concepts are the notions of abstract and concrete objects, hierarchical composition of ordered and unordered objects, sharing of components, and reference links. This model has proven useful for specifying a wide variety of document objects, and is the basis for an implemented document preparation system.