SVM Method and Integrating Pinyin and Dictionary Method for Chinese Spelling Errors Detection

Spelling errors are very common in our daily lives. They sometimes cause serious problems. For example, they appear in government work report. Methods to solve the spelling errors problem based on n-gram language model are widely used. SVM is a famous classifier adopting Hyperplane. This paper introduced a method taking advantage of the SVM model to detect wrong characters. The correct words were considered as positive cases, and the others were considered as negative cases in the training data set. We trained the SVM model and detected wrong characters by the model. Besides, another method is proposed, which integrated the pinyin factor and the dictionary factor to spelling errors detection. Experimental results show that the method based on the SVM model and the method including the pinyin and dictionary factors based on n-gram language model outperform the standard n-gram language model method.