汉语文本读入中音字转换的知识集成模型和时间同步搜索算法

Knowledge-Integrated Model and Time-Synchronously Search Algorithm for Speech-Character Transchiption in Chinese Text Read-in System

本文提出一种集成声学和语言学（特别是词间近邻关联和远邻语法规则）知识的组合概率模型，以及将这一模型用于音字转换过程的时间同步的动态规划搜索算法，并实验验证了所提模型和搜索算法的有效性。

Abstract This paper presents a combined probabilistic model integrating acoustic and linguisticknowledges (especially the link relationship between neighbouring words and the grammatical rules between distant words), and a time-synchronously dynamic programming search algorithm for applying this model to speech-character transcription. The effectiveness of this model and search algorithm has also been tested and verified.