摘要
新型冠状病毒(COVID-19)疫情爆发期间,涌现出了众多的抗疫意见领袖。通过对意见领袖话题传播和演化进行分析研究,可以为网络舆情治理和疫情防控提供理论和知识支撑。采用N-Gram语言模型和Shin⁃gling相似度算法相结合的方式进行话题检测,再通过Neo4j图数据库存储与检索意见领袖、话题、事件等多维实体特征,构建以意见领袖为核心的话题图谱。实验结果表明,话题准确率达82.3%,召回率达81.6%,与传统Single-Pass聚类相似度算法相比均有所提高。通过对图谱分析,能够简单直观地展示出不同实体间多维舆情关系。同时,可以提高检索速度和分析效率,符合舆情传播客观规律。
Many anti-epidemic opinion leaders emerged during the outbreak of COVID-19 period.Through the analysis and research on the topic dissemination and evolution of opinion leaders,it can provide theoretical and knowledge support for network public opinion governance and epidemic prevention and control.This paper first uses the combination of N-Gram language model and shingling simi⁃larity algorithm for topic detection.Then by storing and retrieving the multi-dimensional entity characteristics such as opinion leaders,topics,events and so on,Neo4j graph database is used to build topic graph with opinion leaders as the core.The results show that topic accuracy reaches 82.3%and recall rate 81.6%,which are improved compared with the traditional Single-Pass clustering similarity al⁃gorithm.Through the analysis of the Graph,the multidimensional public opinion relationship between different entities can be displayed simply and intuitively.At the same time,it can improve the retrieval speed and analysis efficiency,and conform to the objective law of public opinion dissemination.
作者
任东亮
林绍福
黄鸿发
付钰
REN Dong-liang;LIN Shao-fu;HUANG Hong-fa;FU Yu(School of Software,Beijing University of Technology,Beijing 100124,China;Beijing Smart City Research Institute,Beijing University of Technology,Beijing 100124,China;TRS Information Technology Co.,Ltd,Beijing 100101,China)
出处
《软件导刊》
2020年第10期20-24,共5页
Software Guide
关键词
新冠疫情
意见领袖
网络舆情
知识图谱
话题分析
COVID-19 epidemic
opinion leader
network public opinion
knowledge graph
topic analysis
