Accurate Braille-Chinese translation towards efficient Chinese input method for blind people

People with visual disabilities have difficulties in inputting Chinese into computer via conventional methods. Although some specialized devices and software can help Chinese blind users input Mandarin Braille into computers, translation of Mandarin Braille into Chinese is still a problem. In this paper, a new approach for translation of prevailing Mandarin Braille to Chinese character is proposed. By using characteristics of prevailing Mandarin Braille such as word segmentation-concatenation rule and tone information in the N-gram decoding process, the translating approach is able to achieve higher performance than current methods. Furthermore, under a paragraph and article translation scheme, language model adaptation is adopted by using combination of N-gram language model and cache-based language mode to improve translation performance. Experimental results show that the overall precision for the translation is 94.3%. If proper nouns are not considered, the precision can reach 95.1%.