%0 Journal Article
%T 一种基于树核函数的半监督关系抽取方法研究<br>A semi-supervised method based on tree kernel for relationship extraction
%A 刘晓勇
%J 山东大学学报（工学版）
%D 2015
%R 10.6040/j.issn.1672-3961.1.2014.259
%X 摘要： 为了解决传统的半监督关系抽取算法易产生的"语义变异"问题,提出一种新的基于树核函数的半监督关系抽取算法。该算法主要采用树核函数和种子集约束扩展两个策略,弱化"语义变异"现象带来的关系抽取不够准确的问题,提高关系识别的正确率。在基准数据集PopBank上的试验研究表明,提出的使用约束机制扩充种子集的半监督学习方法在4个评价指标上(Precision, Recall, F-measure, Accuracy)均优于常用的两种关系抽取方法,从而验证了该算法与其他算法相比能够具有较好的关系抽取能力。<br>Abstract: It was difficult for traditional semi-supervised relation extraction methods to solve "semantic variation" problem. A new semi-supervised relation extraction algorithm based on ensemble learning was prorosed and named L-EC-RE, which used two strategies, one was tree kernel and the other was constrained extension seed set. Experimental study on PopBank benchmark data sets showed that L-EC-RE had better performance than two usual relation extraction algorithms in four assessment criteria, which were Precision, Recall, F-measure and Accuracy
%K 支持向量机
%K 语义变异
%K 树核函数
%K 关系抽取
%K 半监督方法
%K &lt
%K br&gt
%K relationship extraction
%K semi-supervised method
%K semantic variation
%K tree kernel
%K support vector machine
%U http://gxbwk.njournal.sdu.edu.cn/CN/10.6040/j.issn.1672-3961.1.2014.259