Pinyin⁃Chinese character transform is an important aspect of Chinese language information processing, and has a wide range of applications in such aspects as speech recognition, pinyin input, etc. This paper reviews the theoretical knowledge of the word tree, and then solve the word tree, tree pruning last word, with the attendant, to produce multi⁃tone word processing. Based on the above theory, the paper proposes a kind of efficient Pinyin⁃Chinese character transform decoding algorithm based on language model, which is related to a number of the following technical methods as the revaluation of zero probability,path pruning, multi⁃tone word processing and many other technologies. Experimental results show that the algorithm can reach 97. 78% accuracy rate, and compared with the conventional method, the effect is more significant.%音字转换是汉语言信息处理的一个重要方面，在语音识别、汉语拼音输入等方面都有广泛的应用。本文首先回顾了词树的理论知识，然后对词树进行求解，最后对词树进行剪枝，随之，对产生的多音字进行处理。基于上面的理论，提出了一种基于语言模型实现音字转换的高效解码算法，这种算法涉及到零概率重估、路径剪枝和多音字处理等多项技术，实验结果表明这种算法可以达到97．78％的准确率，与以往的算法相比，效果比较显著。
展开▼
