Recognition of Equations Using a 2-D Stochastic Context-Free Grammar

Recognition of printed equations is performed using a two-dimensional stochastic context-free grammar. A stochastic context-free grammar is a context-free grammar (CFG) in which every production rule is labeled by a probability. A two-dimensional context-free grammar is a CFG in which strings are m x n arrays of letters, and vertical concatenation is supported in addition to horizontal concatenation, when the appropriate dimensions match. In this work, a CFG that is both two-dimensional and stochastic is used to model probabilistically the way images of equations are generated. Thus we combine the hard structural constraints provided by grammatical rules with the soft probabilistic constraints provided by stochastic models.