Automatic spelling correction rule extraction and application for spoken-style Korean text

Nowadays, spoken-style text is prevailing because lots of information are being written in spoken-style such as Short-Message-Service(SMS) messages. However, the spoken-style text contains more spelling errors than the traditional written-style text. In this paper, we propose a rule-based spelling correction system which can automatically extract spelling correction rules from the correction corpus and apply extracted rules to spelling errors of input sentences. In order to preserve both high precision and high recall, we devise a candidate-elimination algorithm which determines appropriate context size of spelling correction rules based on rule accuracy. Experimental results showed that the proposed system can extract 42,537 spelling correction rules and apply the rules to correct spelling errors on the test corpus and thus, the rate of precision is increased from 31.08% to 79.04% on the basis of message unit.