电子政务领域中文术语层次关系识别研究

数据驱动下,与日俱增的电子政务信息资源愈发表现出多源异构的特性,基于大规模语料设计一套电子政务领域内中文术语深度层次关系的自动化识别方案,不仅有利于从内容与结构层面弥补人工构建领域词表的不足,且对于我国政务信息资源的开放共享与后续应用更具有重大现实意义。因此,本文分别基于内容与结构双重视角识别电子政务主题词表内术语间的深层关联,通过谱聚类生成的基于内容的层次关系为初步框架,凭借形式概念分析生成的基于结构的层次关系为后期修正指导,以期构成兼顾关联术语召回率与准确率的电子政务领域术语本体。研究结果显示,电子政务术语本体的层次结构合理有效,且术语层次关系的评价结果表明知识本体具备良好的扩展性和延伸性。

Research on Hierarchy Identification of Chinese Terms in the Field of E-government

Driven by data,the characteristics of multi-source heterogeneity have been demonstrated among an increasing number of e-government information resources.Utilizing a large-scale corpus to design an automatic identification scheme for the deep hierarchy of Chinese terms in the field of e-government not only compensates for the lack of a man-made thesaurus in terms of content and structure but also has great practical significance for the dissemination and subsequent application of government information resources in China.Therefore,in line with the dual perspectives of content and structure,the deep associations between terms in the e-government thesaurus are identified in this paper.The content-based hierarchy that is generated by spectral clustering is utilized as the preliminary framework and the structure-based hierarchy that is generated by formal concept analysis is employed as the later modification guide,in order to form the ontology of e-government terms that accounts for the recall and accuracy rates of related terms.The results reveal that the hierarchy of the ontology of e-government terminology is reasonable and effective,and the evaluation results of the term hierarchies illustrate that the knowledge ontology has excellent expansibility and extensibility.