Skip to Main Content
In this paper we propose an implementation method for an off-line layout recognition and semantic understanding system for mathematical formulae. This off-line system aims at higher order coding of mathematical formulae in scientific articles as an application in document analysis. The system has two intermediate output codes: a layout tree, holding information of geometrical structure of the formula and character recognized code of the symbols, and a semantic tree, holding information of semantics of symbols. From the structure tree and the semantic tree after layout recognition and semantic understanding, various useful outputs can be generated at the translating part. This paper mainly describes implementation techniques for LATEX source output for high quality typesetting and gnuplot script output for drawing a function as a method for visual representation.