Model-Based Conversions of LATEX Documents

We are creating a document conversion system based on modelling the logical structures of two broad categories of document types: human-oriented and machine-oriented. Each human-oriented, or user, type is a genre of documents that is well defined and well known to human authors and readers; for example, “scientific articles”, “mathematical formulae” (e.g., x 3+x2−1 x2+1 ), or “limericks”. Each machine-oriented, or agent type, is the set of legal data objects of a document processing tool, for example, “LATEX documents” or “troff/EQN documents”. Our models specify the abstract, rather than the concrete (i.e., surface) syntax, of particular documents. Thus we work with documents as tree data structures, whose concrete presentations have a surface syntax of no intrinsic interest. By so proceeding we hope to best utilize the computational sciences, and sidestep what we believe to be the red herring of “markup language”.