项目名称: 社交网络中基于短文本的事件检测与分析理论及关键技术研究
项目编号: No.61472337
项目类型: 面上项目
立项/批准年度: 2015
项目学科: 自动化技术、计算机技术
项目作者: 李青
作者单位: 香港城市大学深圳研究院
项目金额: 82万元
中文摘要: 社交网络已成为应用普及反应迅速的群智传感网。其中的文本数据来源广、生成快,在短时间内快速地聚集大量语义丰富的信息,从中检测及分析事件将成为实时了解社会发展动态,进而及时形成适当决策的重要渠道之一。但其中大多属于短文本,包含的特征词少、语境信息不全,因此给精准地检测及分析事件带来了极大的挑战。本项目的主要任务及目标包括:1)研究短文本的特征词选择及赋权方法,并利用社交网络用户之间的关系来提高事件检测的精准度;2)基于用户、文本及标签等维度联合构建潜在语义主题模型,全面精准地挖掘事件内容的语义;3)结合时间信息与公众情感构建事件的生命周期模型,以辅助分析事件演化及发生的原因;4)开发一个实验原型系统,利用真实社交网络数据验证本项目提出的理论及技术的有效性。研究成果将填补从短文本中挖掘事件内容语义及分析事件生命周期的空白,并为信息扩散模型的构建、事件检索等相关研究提供新的思路。
中文关键词: 社交网络;群智;短文本分析;事件检测;情感分析
英文摘要: Social networks have become widespread and sensitive crowd-sourcing sensor networks, which accumulate plentiful text with rich semantics. Event detection and analysis from such text has become an important channel for tracking in real time the dynamic status of social development and making correspondingly appropriate decisions. However, most text in social network is short text, hence causes low accuracy of event detection and analysis. In this project, the main objectives and tasks are to: 1) improve the accuracy of event detection by proposing a new term-weighting method for short text, and exploiting the users' relationships in social network; 2) mine the semantics of the events' content comprehensively and accurately by jointly modeling the users, text and labels in the latent topic model; 3) analyze an event's evolution trend by constructing the life cycle model of event based on time and social sentiments; 4) validate the effectiveness of the proposed theories and techniques using the real social network data by developing an experimental prototype system. The research in this project is the first attempt of mining the semantics of event content from short text, in addition to constructing event life cycle model from both the temporal aspect and social sentiments, and will shed a light on new research directions of information diffusion model, event retrieval, etc.
英文关键词: social network;crowdsourcing;short text analysis;event detection;sentiment analysis