使用树核函数进行中文代词消解的研究

Research on tree kernel for Chinese pronominal coreference resolution

指代消解是自然语言处理中语篇理解的关键问题之一。近年来对于树核函数的研究为指代消解提出了新的思路。基于树核函数,提出了一种新的中文代词的消解方法。将句法树作为一个特征,使用SVM提供的树核函数自动获取句法信息。在ACE2005 NWIRE基准数据上进行实验的结果表明树核对中文代词的消解能起到显著的作用,其中F值达到了75.8%。

Coreference resolution plays an important role in natural language processing.Research on tree kernel function in recent years has put forward new means for coreference resolution.This paper proposes a tree kernel-based approach to anaphora resolution of pronoun in Chinese.Convolve kernel of SVM is used to obtain structured information.Then this paper combines syntax tree as a feature with other base features.Evaluation on the ACE2005 NWIRE shows that tree kernel can much improve the performance,especially achieving F-measure of 75.8%.