项目名称: 多模态跨平台社会事件跟踪与预测技术研究
项目编号: No.61303173
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 自动化技术、计算机技术
项目作者: 张天柱
作者单位: 中国科学院自动化研究所
项目金额: 28万元
中文摘要: 随着互联网上网页数量的急剧增长,其新闻事件的有效组织和监控已经成为一个实际的挑战。为了解决这个问题,我们提出了有效的社会事件描述、关联跟踪和全局态势分析与预测方法,从而建立实时、有效、鲁棒的社会事件分析模型。关于社会事件的多源信息融合与建模,我们基于自然语言理解技术和图像视频处理技术,提出了基于语义的多模态信息融合的方法。为了有效地实现社会事件的关联跟踪,我们提出了基于协同学习的方法,从而实现了跨模态、跨平台和跨时空的语义关联,建立了社会事件的语义关联体系。关于社会事件的全局态势分析与预测,我们提出了三种不同的策略对不同的事件进行预测,包括社会事件的主题挖掘、模式预测和统计预测。基于事件预测的发展轨迹,有利于揭示社会事件的传播机制,从而掌控社会事件和网络舆情的发展过程,实现社会事件的监控和预防。
中文关键词: 跨媒体分析;多模态分析;跨平台分析;社会事件预测;社会事件跟踪
英文摘要: With the massive growth of events in Internet, efficient organization and monitoring of events becomes a practical challenge. To solve this problem, we attempt to adopt three steps including an effective social event description, social event tracking, and global trend analysis and forecasting. As a result, we can establish a real-time, efficient and robust social event analysis platform. About the first step for social event multi-source information fusion and modeling, we propose a semantic-based multi-modal fusion method which is realized by natural language processing technology and image and video processing technology. In order to achieve the social event tracking, we propose a collaborative learning method in order to achieve a cross-modal, cross-platform and cross-spatio-temporal semantic association, and obtain the semantic associated system of social events. About the global trend analysis and prediction of social events, we propose three different strategies to predict different events, including social event topic mining, social event pattern prediction and social event statistical forecasting. Based on the results, it is conducive to reveal the mechanisms for the dissemination of social events, which can control the processes of social events and public opinion.
英文关键词: cross-media analysis;multi-modality analysis;cross-platform analysis;social event forecasting;social event tracking