Correcting English Text Using PPM Models

An essential component of many applications in natural language processing is a language modeler able to correct errors in the text being processed. For optical character recognition (OCR), poor scanning quality or extraneous pixels in the image may cause one or more characters to be mis-recognized; while for spelling correction, two characters may be transposed, or a character may be inadvertently inserted or missed out. This paper describes a method for correcting English text using a PPM model. A method that segments words in English text is introduced and is shown to be a signiicant improvement over previously used methods. A similar technique is also applied as a post-processing stage after pages have been recognized by a state-of-the-art commercial OCR system. We show that the accuracy of the OCR system can be increased from 96.3% to 96.9%, a decrease of about 14 errors per page. 1 Motivation In order to fully evaluate the performance of text compression algorithms, large bodies of material have to be made available in machine readable form. Optical character recognition (OCR) systems provide a fast and relatively inexpensive option for acquiring such text. However, when confronted with digitizing over 3,000 pages of Dumas Malone's Jeeerson and his time (Malone, 1977), we found that the time taken to correct the text once it had been processed by an OCR system was prohibitive. Some common mistakes made by the OCR software were the letter c being confused with the letter e (for example, the word thc occurring instead of the, and Jcccrson instead of Jeeerson), problems with the letters m and w, l's being replaced by i's, the mis-recognition of question as guest ion and upper case characters appearing in the middle of words. The task of correcting these errors added substantially to the time required to complete the project. (On average, a further three minutes was required to correct each page.) This paper describes a method of correcting these errors. The next three sections describes the theoretical background to the approach we adopt. After that, we describe how the method can be applied to two speciic problems|automatic segmentation of words in text, and improving OCR output.