Knowledge Integration in Text Recognition

Sargur N. Srihari, Jonathan J. Hull

The paper describes an algorithm based on AI techniques for recognizing words of printed or hand-written text--with the technique developed also applicable to correcting substitution spelling errors. The algorithm effectively integrates bottom-up information in the form of letter shapes, letter transitional probabilities and letter classification-error probabilities together with top-down knowledge in the form of a lexicon of legal words represented as a letter trie. Experimental results with the algorithm are reported for the combined top-down and bottom-up approach and for each of the two approaches individually.


This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.