A Document Analysis Method Based on a Consistent Structural Model of Document Elements and Pages

A document analysis method based on a structural model nf document elements and pages is described. byout analysis for generic dmumcnt elements and logical structure analyaia For upecific documents are integrated in a consistent way. Each category of document element is defined as a "class". Each definitien of class consists of a definition type. classes of subordinate elernenLq with logicnl labels and geometric con- atraints among them. Classes of specific pages can be defined as well as classes of elements. Class definitions form a net- work whose nodes correspond to element classes and whom links correspond to elementrsubordinate reIations. The rec- ognition process storb from the most primitive element class, and traverses the network to find elements satisfying geo- metric constraints in the definition. It progresses sequen- tially from lower element classes Za higher, until all defini- tions have been exemined. A prototype system has been im- plemented. Experimental results Are consistent with results obtained rrom procedure-oriented systems, and they show that the prototype can produce a set of lobtally labeled ele- ments uaing classes or specific document pages.