Issues in Performance Evaluation of Mathematical Notation Recognition Systems

Abstract

Performance evaluation of document recognition systems is a difficult and practically-important problem. In this thesis, we contribute to the understanding of performance evaluation by studying some issues that arise in evaluation of systems for recognition of mathematical expressions. Issues that are discussed cover the reported performance evaluation experiments, the code availability, the nature of the mathematical notation, the extent of the coverage of mathematical recognition systems, and the quantification of performance evaluation results. For each issue, we discuss its impact on performance evaluation, give an overview of the state of the art for addressing it and point out open problems.