摘要
针对中文信息抽取系统中建立提取事件模板的难点问题,基于Bootstrapping思想,提出一种简单、可行的实体关系自动生成方法,利用由种子词和种子模板组成的知识库建立学习器,采用标量聚类的方法,通过种子模板抽取更多的与种子词相似语义关系的特征词。在此基础上,利用最近邻居的原则,进而生成更多的抽取模板。丰富了知识库,为分析二元实体关系奠定基础,使得生成复杂的消息模板成为可能,同时极大地减轻手工建立模板的复杂度,有利于系统进行移植。
A method of Chinese automatic entities relation extraction is proposed in this paper based on Bootstrapping algorithm in order to solve the problem of event template extraction in Information Extraction （IE） systems. This method makes use of seed words and seed patterns to build a learning program, which extracts more characteristic words using Scalar Clusters. These characteristic words have semantic similarity with seed words. Then more extraction patterns could be learned automatically and added to the knowledge database, which is a foundation for analysis of two-entity relation and makes it possible that complex event template could be acquired automatically. This method reduces greatly the working load in manually constructing patterns and makes IE systems more feasible and portable.
出处
《微电子学与计算机》
CSCD
北大核心
2006年第12期15-18,共4页
Microelectronics & Computer
基金
国家863计划重大项目(2001AA114210)
作者简介
张素香（1973-），博士研究生，讲师。研究方向为自然语言理解、机器学习。
