项目名称: 基于概率图的话题传播与演化分析方法研究
项目编号: No.61302144
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 无线电电子学、电信技术
项目作者: 胡艳丽
作者单位: 中国人民解放军国防科学技术大学
项目金额: 28万元
中文摘要: 在网络舆论分析中,追踪话题的动态演化和传播过程具有重要的理论意义和应用价值。由于话题具有隐含性、语义模糊性和不确定性,且舆情信息规模巨大、网络成员间的社会影响复杂多变,如何分析话题传播和演化一直是网络舆论演化分析面临的主要挑战之一。本课题研究基于概率图的话题传播和演化分析方法,旨在建立话题传播和演化分析的统一框架,获取准确、有效的话题演化关系和传播路径。拟开展以下研究:(1)在话题演化方面,针对话题的语义模糊性和不确定性,采用概率生成模型提出在线子话题发现和关联分析方法;(2)在话题演化分析基础上,提出基于贝叶斯网络的话题传播分析方法;因为先验知识对模型求解的准确性具有重要影响,进而提出基于社会网络分析的话题传播先验知识建模方法;(3)在应用方面,针对理论验证困难和实证研究匮乏问题,实时有效采集大规模舆情信息和在线社会网络数据,采用真实的话题传播和演化实例验证上述理论方法的准确性与实用性。
中文关键词: 话题传播与演化;概率图模型;在线社会网络;基于话题的用户影响力评估;信息扩散能力预测
英文摘要: As one of the fundamental problems, the diffusion and evolution of topics plays an important role in the analysis of public opinions in online social networks. Topics are spreading among user-generated documents through online social networks, together with the content evolution by introducing novel contents into documents. Both the diffusion paths and the evolutionary process of a topic are implicit with ambiguousness and uncertainty, making them much challenging to be discovered. Since the diffusion and evolution of topics are tightly interweaved, this project aims to simultaneously track the evolution of any arbitrary topic and reveal the latent diffusion paths of that topic in online social networks based on probabilistic graphical models (PGM). To this end, we intend to conduct research in the following subjects: (1) In the research of topic evolution, inspired by probabilistic generative model due to the ambiguousness and uncertainty of topics, we propose a topic evolution model, consisting of incremental sub-topic detection and correlation analysis; (2) In the research of topic diffusion, we propose methods based on Bayesian networks for revealing the latent diffusion paths of topics, and then model prior knowledge of topic diffusion through network projection taking into consideration of social influence
英文关键词: topic diffusion and evolution;probablistic graphical models;online social networks;topic-based influence ranking;prediction of information diffusion