基于社交媒体网络的事件检测

发布时间：2018-10-13 15:20

【摘要】：当今世界正在见证一个大数据和信息爆炸时代,每天由不同信息源产生的以不同模态呈现的海量新闻报道、社交媒体信息等。当代的新闻消息,不仅仅通过新闻网站、报纸、广播等传统媒体进行报道,也利用丰富多变的社交媒体网络平台(像微博、Twitter、Facebook、Instagram等)进行广泛地传播,快速成为人们谈论的话题。在传统的自然语言处理领域,可以通过提取新闻网站中报道信息,检测新闻事件。但处于以社交媒体为主导的互联网时代中,从单一的官方新闻媒体了解事件的官方信息,更需要从社交媒体网络中了解相关事件下大众,以及事件相关的内容信息。在本文中,我们将以社交媒体为代表的新媒体和以新闻报道为代表的传统媒体相结合,利用快捷、有效、受众广泛的社交媒体信息对检测到的新闻媒体的事件进行分析,提出一种基于社交媒体网络的新闻事件检测与信息增强方法。本文所提框架共包括三部分:数据采集模块、事件检测初始化模块、已检测事件信息增强模块。大规模的抓取了新加坡范围内Twitter数据和新闻数据,构建数据集,运用自然语言处理技术对文本进行预处理,并提取新闻文本中的多种特征属性,提出了结合共享时间、共享关键字、共享社交化标签的事件类摘要的方法,对Twitter数据集进行查询,而后滤除与事件无关的冗余数据,用具有地理标签的相关Twitter数据对已检测的新闻事件进行信息增强,并在地图上对已检测事件相关Tweets增强数据进行展示。
[Abstract]:Today the world is witnessing an era of big data and information explosion, with mass news reports, social media information and so on, produced by different information sources and presented in different modes every day. Contemporary news not only reports through traditional media, such as news websites, newspapers, radio and so on, but also makes use of the rich and changeable social media network platform (such as Weibo, Twitter,Facebook,Instagram and so on) to carry on the widespread dissemination, quickly becomes the topic that people talk about. In the traditional field of natural language processing, news events can be detected by extracting information from news websites. But in the era of the Internet dominated by social media, it is more necessary to know the public and the content of the event from the social media network to understand the official information of the event from a single official news media. In this paper, we will combine the new media represented by social media with the traditional media represented by news reports, and analyze the events of the news media detected by using the quick, effective and wide range of social media information. This paper presents a method of news event detection and information enhancement based on social media network. The proposed framework consists of three parts: data acquisition module, event detection initialization module and detected event information enhancement module. This paper grabs the Twitter data and news data in Singapore on a large scale, constructs the data set, uses the natural language processing technology to preprocess the text, and extracts a variety of characteristic attributes from the news text, and proposes a combination of sharing time. Sharing keywords, sharing event class abstracts with social tags, querying Twitter datasets, filtering out redundant data independent of events, and enhancing information of detected news events with Twitter data with geographical tags. The Tweets enhancement data related to detected events are displayed on the map.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.1

【相似文献】