Survey of Mathematical Expression Recognition for Printed and Handwritten Documents

Abstract

In this paper, we provide a review of Mathematical Expression Recognition (MER) for both printed and handwritten domains. We describe the past work in a manner that clearly highlights the common and distinguishing aspects. The review reveals the advances in the techniques used for the detection and recognition of MEs for online and offline modalities. Systems have evolved from the use of domain-specific crafted modules to architectures learned in a completely data-driven manner. Learning tasks and datasets have been standardized, thus allowing performance comparison of the systems. With automated learning of improved feature representations and guided by the related fields such as machine translation and captioning, the systems have progressed to demonstrate improved robustness to the variability in the inputs.