在线社交网络中地域性话题发现

发布时间：2018-05-21 09:46

本文选题：社交网络 + 地域性话题　；参考：《计算机学报》2017年07期

【摘要】：在线社交网络中日益丰富的地理位置信息为传统舆情感知、信息检索技术带来了新的思考.文中以在线社交平台Twitter为研究对象,以社交网络中地域性话题(Geographical Topic)发现为研究目标,工作主要分为社交网络话题性和地域性分析、地域性话题发现两个部分.首先,文中基于用户、位置和话题间的相互关系,阐述了社交网络用户具有地域性和话题性特征,分析了地理位置和话题对词项使用的影响,抽象出地域和话题之间的关联.其次,根据地域性话题的空间关联特征,综合考虑用户发布的文本内容和地理位置信息,按照主题模型思想构建地域性话题发现模型GTTD(Geographical Textual Topic Discovering model),将用户、话题和地理位置间存在的紧密关系同时引入话题发现框架中.最后利用吉布斯采样算法进行模型的参数估计.基于Twitter真实数据集的实验表明:文中提出的GTTD模型能有效地发现社交网络中的地域性话题,并且与LGTA、Geofolk模型对比,在困惑度(perplexity)指标上体现出优势.
[Abstract]:The increasingly abundant geographical location information in online social networks brings new thinking to traditional public opinion perception and information retrieval technology. In this paper, the online social platform Twitter is taken as the research object, and the geographical topic discovery in the social network is taken as the research goal. The work is mainly divided into two parts: the topic of social network and the regional analysis, and the regional topic discovery. Firstly, based on the relationship among user, location and topic, this paper expounds the regional and topic characteristics of social network users, analyzes the influence of geographical location and topic on the use of words, and abstracts the relationship between region and topic. Secondly, according to the spatial correlation characteristics of regional topic, considering the text content and geographical location information published by users, a regional topic discovery model (GTTD(Geographical Textual Topic Discovering model) is constructed according to the idea of topic model. The close relationship between topic and geographical location is also introduced into the topic discovery framework. Finally, Gibbs sampling algorithm is used to estimate the parameters of the model. Experiments based on the Twitter real data set show that the proposed GTTD model can effectively find the regional topics in social networks, and compared with the LGTA Geofolk model, the proposed GTTD model is superior to the LGTA Geofolk model in terms of the degree of confusion and the index of complexity.
【作者单位】：东南大学计算机科学与工程学院;东南大学计算机网络和信息集成教育部重点实验室(93K-9);
【基金】：国家“八六三”高技术研究发展计划项目基金(2013AA013503) 国家“九七三”重点基础研究发展规划项目基金(2010CB328104) 国家自然科学基金(61272531,61202449,61272054,61370207,61370208,61300024,61320106007,61472081) 江苏省网络与信息安全重点实验室(BM2003201) 江苏省科技计划项目(SBY2014021039-10) 无线通信技术协同创新中心社会公共安全科技协同创新中心资助~~
【分类号】：TP391.1

【相似文献】