A New Approach for Automatic Chinese Spelling Correction

This article presents a new approach for automatic Chinese spelling error detection and correction. Existing Chinese spelling checking systems have two problems: (1) low precision rate, and (2) lack of correction capability. The proposed Chinese spelling correction method is composed of two mechanisms (1) composite confusing character substitution, and (2) advanced word class bigram language model. The characters in the input sentence are rst substituted by their corresponding composite confusing character sets one by one. A composite confusing set is the collection of similar characters to a Chinese character from multiple views of shape, pronunciation, meaning , and input keystroke sequence. The substitution step produces several sentence hypotheses for the input sentence. Then, an advanced word class bigram language model, such as inter-word character bigram (IWCB) or SA-class bigram can be used for scoring each sentence hypothesis. Finally, the best scored sentence hypothesis is compared with the input sentence to determine the typos and their corrections. Experiments show that the proposed approach is very eeec-tive for dealing with the two mentioned problems.