Automatic chinese topic term spelling correction in online Pinyin input

Usually people use the input software to type Chinese language on a computer. The software takes a three-step approach: (1) receive the English keyboard input; (2) convert it to Chinese words; (3) output the words. Traditional Chinese spelling correction algorithms focus on the errors in the output Chinese, but ignore the errors introduced in the original keyboard input. These algorithms do not work well because the errors in the output are usually not the type of the typographical errors, which these algorithms are good at. In this paper, we propose a novel Chinese spelling correction model directly targeting at the original keyboard input. We integrate this model to an online Chinese input method, to improve the spelling suggestion feature. Experiments using real-word data show that this model helps the spelling suggestion achieve a 93.3% accuracy.