CSAIL Event Calendar: Previous Series
Recognition of Handwritten Mathematical Expressions
Speaker: Nick Matsakis , Learning and Vision Group - MIT AI Laboratory
Recent years have seen the availability of a number of commercially successful products which recognize a user's natural handwriting and use this ability to perform simple tasks such as scheduling appointments and writing memos. However, more complex recognition tasks dealing with structured data, such as diagrams, graphs, and equations, are still outside the reach of such systems. Recognition of handwritten mathematics is a natural place to begin to bridge this gap, since it can take advantage of previous work in text recognition, but also begin to address such issues as two-dimensional structure and context-dependence. This talk will describe an on-line approach for converting a handwritten mathematical expression into an equivalent expression in a typesetting command language such as TeX or MathML in real time. The three components of this system are a Gaussian classifier of isolated handwritten symbols, a constrained-search algorithm for partitioning an expression into symbols, and a parser which converts a two-dimensional arrangements of symbols into a typeset expression. The talk will include a demonstration of the system working in conjuction with Mathematica to provide a handwriting-based symbolic expression solver.