Sentence-level Intelligent Chinese Input Method On ANDROID Os

Chinese character input technology, which is a hot topic of Chinese natural language processing research, has broad application use. Research of this topic is to design a rational sentence-level Chinese phonetic symbols input method for mobile devices. Then achieve it on a new smart mobile OS-Android, giving out a new set of solution. This input method has high Chinese phonetic symbol-character conversion accuracy and high intelligence, providing an effective Chinese input method for smart mobile device users. In this paper, mainly discuss the design and the implementation:First, focus on the weakness of mobile devices with limited resources, combined with the current information retrieval and data compression technology. Using Trie-tree, which performs well in the retrieval speed and storage space saving, design a new data storage format for sentence-level input method. So searching Chinese phonetic and Chinese character combinations can be completed in linear time. To solve phonetic segmentation problem, design and use a new method called shortest path segmentation algorithm, which can get all Chinese Phonetic segmentations, and better support lacked and fuzzy phonetics input. Use N-gram language model to do Chinese phonetic symbol-character conversion. Then optimize model with N-Best model, which take up less computational time with little lost in accurate rate. These models make the system can run fast on the mobile platform. Use 30,000 common words as system word library, and use 2007 "People's Daily" as the training corpus to build the model. Chinese phonetic segmentation can get all phonetics, and Chinese phonetic symbol-character conversion accuracy rate is 86.92%.When implement the input method software, the system is coded based on Android IMM / IME (Input Method Manager / Input Method Editor) input method framework, then optimize the codes several times. Based on the above model and algorithms, this paper achieves a sentence-level Chinese input method, which can run smoothly on the mobile platform with relatively limited resources. The input unit is sentence, and support word-level association function, performing high input efficiency. Support Android 1.5 and OMS 1.5 operating system, can be widely used on these platforms. Now, the input method software works stable in the real environment-Dopod Mobile A3288, and passed the China Mobile MMarket's "Testing before commercial release", and be "Commercial release" in the MMarket on May 17th, 2010. Meanwhile, the promotion is also underway in relevant BBS and webs.