Abstract: This paper realizes sentence input method using state space model which integrates Viterbi algorithm into.State space of sentence conversion is described successfully.In state space model insertion operation only needs produce candidate words related with Pinyin node recently inserted and needs not modify Pinyin node and candidate word nodes before.Deletion operation is very simple；it only needs delete the candidate node produced by this Pinyin node.It can get the best sentence by father pointer of candidate node with the biggest probability which points to the end of Pinyin node chain in right indicator.State space model is easier to judge previous node and subsequent node than word grid and is proved effective in sentence conversion by experiment.

Key words: sentence input method, state space model, Viterbi algorithm, candidate node

摘要： 提出融入Viterbi算法的状态空间模型，成功地描述了整句转换的状态空间，实现了整句输入法。这种模型的插入操作，只需对新增的拼音节点与已输入的拼音节点可能产生的所有字词候选节点进行概率累积计算，链接到状态空间模型中；删除操作，只需直接删除与拼音节点相关的候选字词节点；而当到达用户输入边界，可以依赖右指针域指向尾节点的候选字词节点中累积概率最大者的父指针，通过回溯找到最优语句候选。该状态空间模型比起词网格更易判断前驱和后继，实验证明其对整句输入法实现是有效的。

关键词: 整句输入法, 状态空间模型, Viterbi算法, 候选节点