A knowledge-based design for structural analysis of printed mathematical expressions

Abstract

Recognition of Mathematical Expressions (MEs) is a challenging Artificial Intelligence problem as MEs have a complex two dimensional structure. ME recognition involves two stages: Symbol recognition and Structural Analysis. Symbols are recognized in the first stage and spatial relationships like superscript, subscript etc., are determined in the second stage. In this paper, we have focused on structural analysis of printed MEs. For structural analysis, we have proposed a novel ternary tree based representation that captures spatial relationships among the symbols in a given ME. Proposed tree structure has been used for validation of generated ME structure. Structure validation process detects errors based on domain knowledge (mathematics) and the error feedback is used to correct the structure. Therefore, our validation process incorporates an intelligent mechanism to automatically detect and correct the errors. Proposed approach has been tested on an image database of 829 MEs collected from various mathematical documents and experimental results are reported on them.