A dive in white and grey shades of ML and non-ML literature: a multivocal analysis of mathematical expressions

With the advent and advancement of machine learning and deep learning techniques, machine-based recognition systems for mathematical text have captivated the attention of the research community for the last four decades. Mathematical Expression Recognition systems have been identified based on terms of their techniques, approach, dataset, and accuracies. This study majorly targets a rigorous review of both the published form of literature and the least attended literature, i.e., grey literature. Apart from the digital libraries, the papers and other instances of information have been gathered from the grey sources like google patents, archives, technical reports, app stores, etc., culminating in 262 instances. After the heedful filtration imposed on both white and grey literature, the final pool of studies has been investigated for eight formulated research questions. The answers extracted have been analyzed, providing both quantitative and qualitative insights. The analysis and surveys have systematically summed up the potentials of both white and grey shades of literature present on MER and brought exciting extractions out of 155 formal white literature and 107 grey sources. The survey extracts and brings out the highlighting observations after analysis, which sublimates the fact that 52% of grey literature is composed of mobile applications and user interfaces, whereas the published 63% of white data is presently concentrated in 39 different conferences, and the prominent conference is ICDAR (#30). A list of challenges and open issues has been extracted for directing future research dimensions.