Abstract: According to Chinese syntax and Nivre algorithm, this paper presents a deterministic dependency parser, which parses Chinese text with consideration of long-distance dependency. It is difficult to parse long-distance dependency with conventional deterministic dependency analysis method. The improved method parses a sentence deterministically without ignoring long-distance dependency. Support vector machines are applied to identify Chinese dependency. Experimental results show that the method outperforms previous system by 5.32% accuracy. The dependency accuracy achieves 78.30%.

Key words: Chinese dependency analysis, Nivre algorithm, Support Vector Machines(SVMs)

摘要： 依据中文语法的特点，提出了Nivre算法和一种远距离依存关系的确定性中文依存关系解析方法。在中文句子中，有些相互依存的词距离较远，使用传统的确定性解析方法进行解析比较困难。在不忽略远距离依存关系的情况下进行确定性依存关系解析，采用支持向量机识别中文依存关系。实验结果表明，依存关系解析精度达到78.30%，提高了5.32%。

关键词: 中文依存关系解析, Nivre算法, 支持向量机

CLC Number: 

• TP391.4