Research On The Key Technology Of Intelligentinput For Communication Terminal

The coming of 3G era, mobile communication terminals are rapidly developed in China. By contrast, we can find the input technologies of Communication terminals which can be described as "foreign codes galloping", while the Chinese searcher need to rely on foreign core technologies. Nowadays, people frequently exchanged their text message or word information. It is essential and imminent to develop a fast and efficient Chinese and English search engine for mobile communication terminal. This paper relied on "application software technology industry and resource sharing services for communication terminal" which is Innovation Fund Projects, and took the embedded intelligent predictive input method reaserch and implementation as the main content, including system kernel, string encoding, and data structure, to solve the current input method's problem of computational complexity high and recognition rate low. For shortcomings of existing Statistical language model, paper established a mixed portfolio of sub-word language model. On this new model, paper presented a multi-classification based on neural network dynamic search algorithms, and used the quality evaluation system to conduct a comprehensive assessment for this algorithm.Because of slowly entering and poor continuity of thinking proposes, Paper proposed a mixed input language model for mobile communication device. Compared with the present language model, the new model joined the word combination characteristics, and used the full-keyboard input way to improve the predict power of statistical language model. The model used first-order Markov probability to estimate the mutual information between adjacent characters which extracted from the known words, and predicted the next character.According to this new language mode, paper established an encrypted program and designed a multi-classification dynamic search algorithm based on neural network. The algorithm solved the problem of large computational complexity by setting the weight and the threshold. Also the algorithm assigned each word with two logos which is historical record and customary experience. It can fastly predict some common words.Finally paper completed the relevant code to implement the algorithm on Mobile 6.0, and successfully transplanted to the Dopod D900 terminal device. Compared with other search engines, paper evaluated the new algorithm with other two algorithms by three important indicators in the light of the State Information Processing Standards GB/T18031-2000. The result showed the multi-hierarchical dynamic search algorithm based on neural network could effectively reduce the number of enters and improve the input efficiency.