中文摘要：

提出了一种利用TBL算法和句法特征进行韵律边界预测的方法.选择语法词的词性、词长和其在语法树中所处的层级这三个句法特征,利用预定义的规则模板,采用基于转换的错误驱动学习算法（Transformation-based error driven learning algorithm,TBL算法）,实现了对输入文本的韵律词和韵律短语的预测.结果表明,韵律词的预测精度达到了98.4%,韵律短语的预测精度达到了82.7%,比其他方法具有更高的预测精度.

英文摘要：

This paper proposes a novel method for prosodic boundary prediction by using syntactic features with TBL algorithms.In this paper,the syntactic features including the part-of-speech（POS）,the length of lexical word and the height of syntax tree are selected to predict the prosodic word boundary and prosodic phrase boundary by using a transformation-based error-drivenlearning（TBL） algorithm.The experiments demonstrate that our method achieves 98.4% of prediction precision on prosodic word and 82.7% prediction precision on prosodic phrase,better than other methods.