Transformation of structured documents

SUMMARY Many documents have a definable structure. Some document formatting systems, like the LaTeX formatter, use a structural notation. In recent years the general mark-up language SGML has gained popularity. In this work we study the transformation of a structure to another. For example, technical journals have their structure definitions, and an article originally written for one journal must be restructured before it can be submitted to another journal. We assume that structure definitions are grammatical, and study what kind of transformations can be automatized or at least semiautomatized. We took a collection of computer science journals and compared their structure definitions. We classified differences as simple, local and global. As transformation techniques we studied syntax directed translation schemata and tree transducers. Our conclusion was that simple and local transformations can be automatized or semiautomatized, depending whether additional information is not needed, while global transformations are difficult to automatize. Transformations were tested in our prototype syntax-directed document processing system. The system has one module for editing a document under one structure definition, and another module for changing a document from one structure definition to another.