网络用户自描述标签向量生成及标签层次体系构建方法  

A method to generate hashtag vector and build hierarchy architecture of network users' self-describing hashtag

作　　者：孙毅 裘杭萍 王沁雪 Sun Yi;Qiu Hangping;Wang Qinxue(Institute of Command and Control Engineering,Army Engineering University of PLA,Nanjing 210000,China;Laiwu Experimental Middle School,Laiwu 271100,China)

机构地区：[1]中国人民解放军陆军工程大学指挥控制工程学院,江苏南京210000 [2]山东省莱芜市实验中学,山东莱芜271100

出　　处：《信息技术与网络安全》2018年第11期44-49,共6页Information Technology and Network Security

基　　金：江苏省自然科学基金(BK20150721;BK20161469);江苏省重点研发计划(BE2015728;BE2016904;BE2017616)

摘　　要：通过用户自描述标签可以快速、高效地对网络用户进行检索、分类和智能推荐。为更好地研究网络用户自描述标签的内在关系,提出了一种网络用户自描述标签层次体系的构建方法。通过对低频标签的理解方法定义了标签向量生成规则。利用搜索引擎语料扩充的方法提取上下位关系的关键模式,实现了一种基于多特征的标签对上下位关系检测方法。以标签频率作为权重,介绍了构建网络用户自描述标签层次树的思路。实验结果表明,本文提出的上下位关系检测方法在精确度、查准率、查全率和F1等指标上较以往的相关工作有大幅度提升。Network users can be quickly,efficiently retrieved,classified and intelligently recommended by using user self-describing hashtags.To study the internal relationship of network user self-describing hashtags better,a method for constructing the self-describing hashtag hierarchy architecture was proposed.The rule of hashtag vector generation was defined by understanding low frequency hashtags. Key modes of hyponymy relations was extracted by using search engine to extend corpus of hashtag pairs.A method to detect hyponymy relationship based on multiple features was proposed.Taking the frequency of the hashtag as the weight,learning from Huffman Tree′s construction idea,the idea of constructing the network users self-describing hashtag tree was proposed.The eperimental results show that the method of detecting hyponymy relationship proposed in this paper has greatly improved compared with the related work in the past in terms of accuracy,precision,recall and F1-score.

关 键 词：用户标签 标签向量 层次体系 上下位关系识别 随机森林 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]