Intelligent Combination of Structural Analysis Algorithms: Application to Mathematical Expression Recognition
 Publication Year:
 2014

 Bepress 193

 Bepress 56
 Repository URL:
 http://scholarworks.rit.edu/theses/7874
 Author(s):
 Tags:
 Gradient based learning; Graph transformer network; Math recognition; Structural pattern recognition
thesis / dissertation description
Structural analysis is an important step in many document based recognition problem. Structural analysis is performed to associate elements in a document and assign meaning to their association. Handwritten mathematical expression recognition is one such problemwhich has been studied and researched for long. Many techniques have been researched to build a system that produce high performance mathematical expression recognition. We have presented a novel method to combine multiple structural recognition algorithms inwhich the combined result shows better performance than each individual recognition algorithms. In our experiment we have applied our method to combine multiple mathematical expression recognition parsers called DRACULAE. We have used Graph Transformation Network (GTN) which is a network of function based systems in which each system takes graphs as input, apply function and produces a graph as output. GTN is used to combine multiple DRACULAE parsers and its parameter are tuned using gradient based learning.It has been shown that such a combination method can be used to accentuate the strength of individual algorithms in combination to produce better combination result which higher recognition performance. In our experiment we were able to obtain a highest recognition rate of 74% as compared to best recognition result of 70% from individual DRACULAE parsers. Our experiment also resulted into a maximum of 20% reduction of parent recognition errors and maximum 37% reduction in relation recognition errors between symbols in expressions.