摘要
实体识别是根据记录所包含的各种描述信息来确定与之相对应的现实实体,记录的相似匹配是数据集成中最 具挑战的工作。分析了常见实体识别算法,提出了实体识别过程框架,用以实现数据ETL的数据规约功能。在开发的实现 语义数据集成的联通统一客户资料系统(UCIS)中,用实体识别算法进行测试,得到的平均返回率和精度分别为86.3%、 96.5%,能够满足工程应用的要求。
Entity identification is to conform the corresponding practical entity based on its various description information. The biggest challenge for data integration is to match the similar records. General entity identification algorithms are analyzed. The entity identification process frame for carrying out the data filtering function of data ETL (data extraction, transformation and loading) is brought forward. Average returning rate and precision tested with entity identification algorithms are respectively 86.3%, 96.5% in developed UCIS(UniCom Client Information System) that carried out data integration based on semantics, which can meet engineering application.
出处
《现代电子技术》
2005年第7期44-46,共3页
Modern Electronics Technique
关键词
数据ETL
相似重复记录
实体识别算法
实体识别过程框架
data ETL
duplicate records
entity identification algorithms
entity identification process frame