Complementary Approaches to Representing Differences Between Structured Documents

Structured documents Documents can be represented as structures with a hierarchical arrangement of text and non-text nodes, where nodes are labelled by category names such as “paragraph” and “section”. Representing documents this way is a natural consequence of using the Standard Generalized Markup Language (SGML) to encode the content and form of documents [10, 11, 7]. SGML is widely used. HTML, the encoding used for World Wide Web documents, is an application of SGML [6]; although HTML is used to build hypertext networks of documents rather than hierarchies, each document is itself a hierarchy with explicitly coded links to build the network. The Text Encoding Initiative uses SGML to encode complex texts [21, 4, 2]. Even documents that are not simple hierarchies can be represented using SGML [3]. Formally, documents represented in SGML or HTML are trees with labelled nodes where the left to right ordering of the offspring of a node is significant. Any piece of text in the document can be treated as a single labelled node, all leaf nodes are text nodes (or empty) and any node with children is a structural or non-text node.