项目名称: 面向海量广播电视监管的新闻拆条与编目技术研究
项目编号: No.61202326
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 计算机科学学科
项目作者: 冯柏岚
作者单位: 中国科学院自动化研究所
项目金额: 21万元
中文摘要: 随着我国监测事业的不断发展以及国内外广播电视新闻数据的海量增长,使得人们对于广播电视新闻资源的监测和管理技术研究提出了迫切需求,特别是从新闻视频结构分析和新闻视频内容理解角度来处理新闻拆条与编目技术变得十分重要。本项目旨在研究面向海量广播电视监管的新闻拆条与编目技术,建立统一的技术框架,并针对其中的关键科学问题展开深入探讨。通过抽取广播电视新闻的结构化特征和融合多模态信息,并结合基于风格描述的新闻条目学习技术,实现对广播电视新闻的结构分析与数据处理;通过利用基于语义要素分析的新闻编目技术,结合静态略览与动态梗概并存的跨媒体编目体系,实现对广播电视新闻的内容理解与监测管理;通过研究基于结构完整和主题一致的新闻编目评价机制,完善对新闻监管技术的理论支持;综合前面各项研究成果,研制一个面向广播电视监管的有效、鲁棒的新闻拆条与编目原型系统,对研究内容进行可行性验证。
中文关键词: 新闻拆条;新闻编目;新闻监管平台;;
英文摘要: With the rapid development of supervision service and the ever-increasing growth of broadcast news data at home and abroad, how to effectively and efficiently supervise and manage these increasing broadcast resources has become urgent, especially video structure parsing and content analysis techniques are considered. In this project, we mainly focus on the techniques of news story segmentation and news summarization in large broadcast video to support a variety of supervision-oriented functions. On one hand, structural feature extraction and multi-modal information fusion techniques, together with style learning based story boundary detection are proposed to achieve the structural analysis and process of broadcast news. On the other hand, semantic element analysis based news summarization, together with video summary and skimming techniques are proposed to achieve the content understanding and supervision of broadcast news. Parallel to this, a summarization evaluation mechanism is considered to further improve the theoretical support of news supervision. Finally, by integrating above research achievements, a novel prototype system of news story segmentation and summarization is designed to verify the feasibility of our ideas.
英文关键词: news story segmentation;news summarization;supervision-oriented news management platform;;