Zhelezniakov D. Methods for handwritten mathematical expression recognition based on machine learning and context-free grammars under constraints

The language of mathematical expressions is an essential part of many domains, including education, engineering, and science. It is a universal language used and un-derstood all over the world. However, creating digital documents that contain complex mathematical expressions is considerably more difficult than creating documents that contain simple text. This is due to the more complex two-dimensional structure of math¬ematical expressions and a large of mathematical symbols alphabet. Usually, mathe¬matical expressions input requires the user to know special mathematical notations or complex editors with many elements and tangled navigation. However, a pen or stylus is the most natural interface for inputting such information. At the same time, the pen¬based interface supports the input of the initial information and provides an ability to change existing electronic documents in the same natural way. Until recently, mainly mobile devices such as smartphones and tablets were equipped with an electronic pen. Nowadays, new types of devices such as interactive panels, digital pens, and smart writ¬ing surfaces have become widely adopted in offices and educational institutions, open¬ing up new opportunities for technologies for recognizing specific handwritten content such as mathematics, diagrams, charts, tables, sketches, etc.