摘要
实体关系抽取是信息抽取的一项重要内容,通过实体关系的抽取能够发现文本中的有价值信息。本文在分析和比较了有监督、无监督、弱监督以及开放式等关系抽取方法的原理和特点的基础上,建立了基于文献的地质实体关系抽取模型:采用统计语言模型作为关系抽取方式、采用Bootstrapping算法作为关系扩展方式。最后据此进行了关联关系发现和关系扩展发现实验。
Relation extraction is an important section of information extraction,which play an crucial role in valuable information discovering.On the ground of analyzing and comparing,including supervised methods,unsupervised methods,self-supervise methods and open information extraction methods,this essay has built a Geologic Entity Relation Extraction Model,using statistical language models for relation extraction and bootstrapping models for relation extension.Finally,according to the above analysis,the experiment of incidence relation discovery and relation extension discovery were carried out.
出处
《中国矿业》
北大核心
2017年第10期167-172,共6页
China Mining Magazine
基金
国土资源部公益性行业科研专项项目资助(编号:201511079)
国家重点研发计划"基于‘地质云’平台的深部找矿知识挖掘"资助(编号:2016YFC0600510)
关键词
文献
关系抽取
统计语言模型
BOOTSTRAPPING
literature
relation extraction
metallogenic prognosis
statistical language model
bootstrapping model