Evaluation of approximate pattern matching algorithms for OCR texts

In recent years there has been going on a large process of digitising old books, articles and
newspapers. These documents are scanned and then processed with Optical Character Recognition
(OCR) software to obtain their text equivalent. However, due to the (usually) poor quality of the
original papers, the OCR software produces text which is not 100% accurate. A simple search for a
pattern in the resulting text would only retrieve those occurrences that were accurately interpreted, but
will ignore incorrectly spelled or distorted variations. In this paper we make use of the recently
devised algorithm by Christodoulakis and Brey (2008), on the edit distance with combinations and
splits, to perform approximate pattern matching for OCR texts. We then compare its performance
against classic generalpurpose
approximate matching algorithms.

Citation:
Christodoulakis, M., Brey, G., Uppal, R.A. (2009) ‘Evaluation of approximate pattern matching algorithms for OCR texts’ Proceedings of Advances in Computing and Technology, (AC&T) The School of Computing and Technology 4th Annual Conference, University of East London, pp.35-42.