An expert system for more efficiently and accurately deducing document structure from document formatting, the expert system including a conversion engine for converting an unstructured file to a structured file, and a verification engine, responsive to the output of the conversion engine, for generating and displaying a representation of the structured file annotated with a visual depictions of the classified components thereof so that the annotations can be modified and/or classifications can be added and/or classifications can be suggested, and/or rules for classification can be suggested and the structured file reprocessed by the conversion engine.