Graph technology applied to editing structured natural-language documents

This is a report on our experiences of applying graph technology in the application area of creating and maintaining structured digital documents during the last four years. Our aim is to contribute to the development of tools for authors of documents with an inherent content structure, such as scientific articles or textbooks, that help building and maintaining that structure. One goal of structural integrity is, for example, referring to terms only after they have been defined and then referring to them consistently with the definition. Another goal is for each section and subsection to play a clear and discernable role in the document. The data from the document we need to help achieve these goals are the location and content of definitions, as well as the location of references. More data may be used to offer more functionality, for example the intended style of a presentation (inductive or deductive as prime alternatives), which would have to be consistent with the relationships of terms and the order of their occurrences in the text. We call the general structure of content in the document the content structure. In the aforementioned form, it consists of concepts that may have various relationships among each other, for example one being a part of another, or being required for the understanding of another concept. Elements of the content structure may be connected to the visible parts of the document, for example as being defined or referred to. We call these visible parts the presentation. This corresponds to the level current word processors regard a document: a hierarchy of sections with typographically marked-up text and other media elements. We call these constituent parts the document hierarchy of divisions on the one hand and the media content on the other. Our tools provide means to construct documents containing content structure as well as the presentation and to check them according to rules of readable structuring. They are to provide technology, not policy in the sense that the author should be free to work with or without content-structural support solely at his discretion. Divisions, concepts, definitions, references and concept relationships naturally form a graph. We have therefore used a graph database and a graph-based programming