基于融合特征相似度的实体消歧方法研究  

Research on entity disambiguation method based on fusion feature similarity

作　　者：张雄[1] 陈福才[1] 黄瑞阳[1] 

机构地区：[1]国家数字交换系统工程技术研究中心,郑州450002

出　　处：《计算机应用研究》2017年第2期347-350,396,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(61171108);国家"973"计划资助项目(2012CB315901;2012CB315905);国家科技支撑计划资助项目(2014BAH30B01)

摘　　要：在数字资源日益丰富的环境下,人名歧义现象为数据检索带来了很多不确定性,降低了数据检索的准确度。运用人名实体的个人信息特征和作者文献话题等特征,采用多特征的融合方法充分挖掘与实体相关联的信息,实现人名消歧。实验使用爬取的某数据库文献数据,进行了人名消歧特征的选取和参数的确定等实验,结果表明提出的方法具有较好的性能。Under the environment of digital resource increasingly rich, personal name entity ambiguity phenomenon has brought a lot of uncertainty for data retrieval, and reduces the accuracy of the data retrieval. This paper used personal charac- teristics of personal name entity and subject characteristics of the literature, fully mined the information associated with the en- tity based on many characteristics fusion method, to achieve the personal name entity disambiguation. The experiments used the crawled literature data of a database, and researched on the feature selection of personal name entity disambiguation and determination of the parameters. The results show that the proposed method has good feasibility.

关 键 词：实体特征 主题信息 特征相似度 人名消歧 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]