MST-based Visual Parsing of Online Handwritten Mathematical Expressions
Lei Hu, Richard Zanibbi · 2016
We develop a Maximum Spanning Tree (MST) based parser using Edmonds' algorithm, which extracts an MST from a directed Line-of-Sight graph in two passes. First, symbols are segmented by grouping input strokes, and then symbols and symbol pair spatial relationships are labeled. The time complexity of our MST-based parsing is lower than the time complexity of CYK parsing with 2-D Context-Free grammars. Also, our MST-based parser obtains higher formula structure and expression rates than published techniques using CYK parsing when starting from valid symbols. This parsing technique could be extended to include n-grams or other language constraints, and might be used for other notations.