信息抽取中实体关系模式的可信度评估

Evaluating the Confidence of Relation Pattern in Information Extraction

在基于Bootstrap的信息抽取技术中,为提高实体关系抽取模式的质量,需要对抽取模式的可信度进行评估。本文提出了根据模式的历史匹配记录来对其进行可信度评估的简单方法,并以此为基础对模式进行了优化合并。经过可信度评估的模式在对句子进行实体关系标注时,有效提高了标注的准确率。这说明该方法能够提高抽取模式的质量,对信息抽取系统的性能提高有一定价值。

In the information extraction technology based on Bootstrap method,in order to improve the quality of the relation extraction pattern,it＇s necessary to evaluate the confidence of the extraction pattern. This paper introduces a simple method of evaluating confidence by pattern＇s matching record,and based on this,the patterns are optimized and merged. When these patterns are used to label the relations of the sentences,the labeling accuracy is effectively increased. So,this method can increase the quality of the extraction pattern,and it＇s of some value in improving the performance of the information extraction system.