基于树核函数的代词指代消解

Tree Kernel Function-based Pronoun Coreference Resolution

提出一种基于树核的英文代词消解方法。针对结构化信息在指代消解中的重要作用,使用SVM提供的卷积树核函数自动获取句法结构信息,将句法树作为一个特征与其他基本特征结合。通过应用不同的剪枝策略,考虑不同句法树对系统的影响,在原有的句法树上扩充一些语义节点。在ACE2004NWIRE基准数据上进行实验的结果证明,该方法对代词的消解起到明显的作用,综合值f提高了11.9%。

This paper proposes a tree kemel-based approach to anaphora resolution of pronoun. On the basis of structured information automatically captured by convolve kernel of SVM, it integrates syntax tree as a feature with other base features. Different pruning strategies are applied to eliminate the impact of syntax trees to the results. Evaluation on the ACE2004 NWIRE benchmark corpus shows that tree kernel can improve thefpefformance by 11.9%. Based on the system, it combines with semantic role feature and verb-driving feature which are acquired from ASSERT.