Typing mathematical queries, in general, can be a demanding task. Mathematical notation is two dimensional and cannot be easily typed with a standard QWERTY keyboard. In order to have an efficient math-aware search system that is capable of locating mathematical expressions and formulas, users should be able to form their queries with the most intuitive and natural method. Handwriting appears to be the most intuitive and promising method to express mathematical queries. Recognition technology for handwritten mathematical notation is far from mature and has never been applied to query generation in a mathematical information retrieval system. Only few methods have been proposed to detect or recover the recognition errors. Recognition techniques for mathematical handwritten notation produce many misrecognized or completely unrecognized symbols. These lexical errors can considerably reduce the performance of a math search system. The objective of this research is to design, implement, and study an automated lexical error compensation system for the query processing front-end of a mathematical IR system which receives pre-processed handwritten mathematical queries. To achieve this objective, a query processing program has been implemented which modifies and expands a handwritten query under different policies, in order to reduce the effect of potential lexical errors on the performance of search. Experiments conducted in this research have demonstrated that these policies can substantially improve the performance of a handwritten-based math IR system.
展开▼
