Word recognition by morphological analysis

The paper is aimed at proposing a correction method for words when an OCR (optical character recognition) engine misrecognizes some characters of the text, or when it does not recognize any characters of the text. When the characters supposed to be misrecognized appear, the candidate characters replace them. Korean morphological analysis is then conducted for the words. If it is successful, the candidate characters are assumed to be correct. If not, one performs Korean morphological analysis for the words with original characters which are misrecognized. If it is successful, the original characters are regarded as correct. In the case of unrecognized characters, they are divided into two categories: one is a partially-recognized character and the other is an unrecognized character. For the former, the possible characters replace them. Korean morphological analysis is then done for the words. If some words are successful, they are regarded as the correct ones. For the latter case, the characters having a high entropy value which is successfully done for morphological analysis, might be the candidates