基于奇异值分解的专利术语层次关系解析研究

发布时间：2019-02-14 11:07

【摘要】：奇异值分解是矩阵分析中常用的分解技术,在高维数据的降维、去噪方面有着广泛的应用。本文将矩阵的奇异值分解技术应用到领域专利术语的层次关系解析中,其核心思想是术语语义空间的深度转换,旨在用某种潜在特征代替文档作为属性来表示术语。具体研究包括,基于位置加权的原始术语-文档语义空间的构建方法、基于奇异值分解的术语特征抽取方法、基于术语-特征语义空间的术语层次关系解析方法。本文对论述的理论方法进行了实证研究,证实了该方法的可行性和有效性,并在此基础上实现较大规模的钢铁冶金领域专利术语层次关系解析,将解析结果纳入到本体学习理论体系中,实现该领域专利术语的知识本体化与可视化。
[Abstract]:Singular value decomposition (SVD) is a commonly used decomposition technique in matrix analysis. It is widely used in dimensionality reduction and denoising of high dimensional data. In this paper, the singular value decomposition (SVD) technique of matrices is applied to the hierarchical relationship analysis of domain patent terms. Its core idea is the depth transformation of the semantic space of terms, which is intended to represent the terms with some potential features instead of documents as attributes. The research includes the construction method of the original term-document semantic space based on position weighting, the term feature extraction method based on singular value decomposition, and the term hierarchy relationship analysis method based on the term-feature semantic space. This paper makes an empirical study on the theory and method discussed in this paper, which proves the feasibility and effectiveness of the method, and on this basis realizes the analysis of the hierarchical relationship of the patent terms in the field of iron and steel metallurgy on a large scale. The analytical results are incorporated into the ontology learning theory system to realize ontology and visualization of patent terms in this field.
【作者单位】：南京大学信息管理学院;南京大学江苏省数据工程与知识服务重点实验室;南京大学计算机软件新技术国家重点实验室;
【基金】：江苏省自然科学基金项目“面向专利预警的中文本体学习研究”(BK20130587) 江苏省“333”工程项目“面向知识服务的中文本体学习研究”(BRA2015401) 国家社科重大招标项目“面向突发事件应急决策的快速响应情报体系研究”(13&ZD174)
【分类号】：TP391.1

【相似文献】