摘要
很多领域都面临实体识别问题,但现有解决框架缺乏通用性。提出了一种基于属性模式的领域无关的实体识别框架。属性的模式代表属性与实体的一种关系,将模式分为四种类型分别处理,针对类型特点提出了更为通用的相似度计算方法。系统根据模式类型决定相似度计算策略,使系统具有更强的扩展性。该框架可以有效综合利用各类属性的特点进行实体识别,结果优于一般的基于属性特征或基于实体关系的方法。
Entity resolution problem involves wide range of fields, nevertheless the existing frameworks lack versatility. A domain - independent entity recognition framework based on the attribute pattern is proposed in the paper. Attribute pattern represents the relationship between entity and attribute, which is divided into four types. The paper proposes a more general similarity calculation method against attribute pattern features. Simiianty calculation strategy is determined according to attributes＇ pattern and format automatically, so the system has strong scalability. The framework can effectively utilize characteristics of various types of attributes, and it is proved more efficient than any approach based on attributes characteristics or based on entity relationship.
出处
《智能计算机与应用》
2014年第1期65-68,共4页
Intelligent Computer and Applications
基金
国家自然科学基金(61003046)
关键词
实体识别
属性模式
扩展性
框架
Entity Resolution
Attribute Patterns
Scalability
Framework
