面向海量新闻简报的事件演变轨迹构建方法  

Constructing Method of Events Evolution Trajectory for Massive News Bulletins

作　　者：赵华 邓志宏 陈英武[1] ZHAO Hua;DENG Zhihong;CHEN Yingwu(College of System Engineering,National University of Defense Technology,Changsha 410073,China;Department of Education and Training,National Defense University,Beijing 100091,China)

机构地区：[1]国防科技大学系统工程学院,长沙410073 [2]国防大学教育与训练部,北京100091

出　　处：《计算机工程与应用》2022年第24期116-124,共9页Computer Engineering and Applications

基　　金：国家自然科学基金(71701205)。

摘　　要：在故事森林模型基础上提出了事件演变轨迹模型。对比故事森林模型,事件演变轨迹模型具备在涵盖不同主题和包含大量冗余信息的海量新闻简报中准确、快速提取热点事件的能力。同时,在分析和理解上下文语义的基础上,以实时在线的方式描述事件的演变过程,在不破坏已有演变结构的基础上,确保用户的浏览体验。基于32 GB的中文新闻简报数据对模型进行了评估和试点用户体验研究。结果表明,事件演变轨迹在准确识别热点事件并将新闻简报组织成符合读者阅读逻辑方面具有更强的能力。Based on the story forest model,the event evolution trajectory model is proposed.Comparing to story forest model,the event evolution trajectory model has the ability to accurately and quickly extract hot-spot events in a massive news bulletin that covers a wide range of topics and contains a wealth of redundant information.At the same time,on the basis of analyzing and understanding contextual semantics,the evolution of events is described in a real-time online manner,ensuring the user’s browsing experience without destroying the existing evolutionary structure.Finally,based on 32 GB of Chinese news bulletin data,the paper evaluates the model and pilots the user experience study.The results show that the event evolution trajectory has more ability in accurately identifying hot events and organizing news bulletins into reading logic.

关 键 词：事件演变轨迹 新闻简报 热点事件 上下文语义 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]