基于地理位置信息的中文微博突发话题检测技术研究

发布时间：2018-10-07 19:49

【摘要】：随着移动互联网的深入发展,基于位置服务(LocationBased Service, LBS)越来越贴近人们的生活,使得微博空间中携带地理位置信息的微博比例越来越高。如何结合地理位置信息,从繁杂的微博空间中检测出突发话题,成为社交网络突发话题检测领域的一个新方向。研究发现,当前结合LBS的社交网络突发话题检测技术尚处于初级阶段。本文在已有研究的基础上,重点分析结合地理位置信息的突发话题检测技术。针对目前微博数据采集技术复杂、低效,现存微博数据集容量小、数据旧,且不包含地理位置信息等问题,在已有微博数据采集方法的基础上,本文提出一种基于应用集调度策略的中文微博数据采集方法,该方法能有效突破微博平台对API接口访问频次的限制,使微博数据采集更加简单、高效。通过对微博用户关系、词语权重、用户影响力以及突发话题检测模型的深入研究,针对目前中文微博突发话题检测方法未能充分利用移动互联网环境下微博平台产生的地理位置信息的问题,本文定义“亲历度”概念,表征微博用户与用户所发微博内容描述事件的亲历程度,提升亲历度高的微博用户话语权;本文提出一种结合地理位置信息的特征词权重计算方法,提高特征词抽取准确率,从而达到提高突发话题检测准确率的目的。在凝聚式聚类算法的具体实现中,存储计算中间值以避免大量重复的查询、计算过程,从而提高聚类算法执行效率。最后,通过实验证明了本文提出的基于地理位置信息的突发话题检测方法的有效性,为突发话题检测研究提供一些新角度。
[Abstract]:With the development of mobile Internet, location based service (LocationBased Service, LBS) is becoming more and more close to people's life, which makes Weibo carry more and more geographical location information in Weibo space. How to detect burst topic from the complicated space of Weibo is a new direction in the field of social network sudden topic detection. It is found that the detection technology of social network burst topic based on LBS is still in its infancy. Based on the existing research, this paper focuses on the analysis of burst topic detection technology combined with geographical location information. Aiming at the problems of Weibo's data collection technology is complex, inefficient, the existing Weibo data set capacity is small, the data is old, and does not contain the geographical position information and so on. This paper presents a Chinese Weibo data acquisition method based on the application set scheduling strategy. This method can effectively break through the limitation of the access frequency of the API interface and make Weibo data acquisition simpler and more efficient. Through the in-depth study of Weibo user relationship, word weight, user influence and burst topic detection model, In view of the problem that Chinese Weibo's method of detecting unexpected topics has not fully utilized the geographical location information generated by Weibo platform under the mobile Internet environment, this paper defines the concept of "degree of experience". This paper presents a method of calculating the weight of feature words combined with geographical information to improve the accuracy of feature extraction. In order to achieve the purpose of improving the accuracy of burst topic detection. In the concrete implementation of the condensed clustering algorithm, the intermediate value is stored to avoid a large number of repeated queries and the calculation process, thus improving the execution efficiency of the clustering algorithm. Finally, the effectiveness of the proposed method based on geographical location information is proved by experiments, which provides some new perspectives for the research of burst topic detection.
【学位授予单位】：哈尔滨工程大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.1

【相似文献】