文献类型：期刊文献

中文题名：中文事件抽取研究文献之算法效果分析

英文题名：Effect Analysis of Chinese Event Extraction Method Based on Literatures

作者：吉久明[1];陈锦辉[1];李楠[1];孙济庆[1]

机构：[1]华东理工大学科技信息研究所,上海200237

年份：2015

卷号：35

期号：12

起止页码：3

中文期刊名：现代情报

外文期刊名：Journal of Modern Information

收录：CSTPCD;;国家哲学社会科学学术期刊数据库;北大核心:【北大核心2014】；CSSCI:【CSSCI_E2014_2016】；

基金：国家社科基金资助项目"面向知识服务的学科领域术语语义分析及应用研究"(项目编号:13BTQ053)的研究成果之一

语种：中文

中文关键词：中文事件;事件抽取;信息抽取;统计分析;情报分析

外文关键词：chinese event; event extraction; information extraction; statistical analysis; information analysis;

摘要：事件抽取是指识别文本中描述在某个时间(或时间段),某个地点或地区,由一个或多个角色参与的某动作的事件。首先对我国关于事件抽取研究的文献进行了总结,给出事件抽取的主要方法及模型。并针对文献中对这些事件抽取方法的效果进行统计分析,探讨各种事件抽取方法或模型的效果及适用性。经过对现有研究文献的统计,结论为:当前有关事件抽取的研究仍在继续,主要集中于金融资讯、会议信息、突发事件、个人简历等来自网页、微博微信等自媒体信息或军事法律等专业文件的事件抽取,所采用的算法包括SVM、CRF、ME、模式匹配、聚类算法等;CRF算法应用与个人简历事件抽取效果最好,采用模式匹配算法的有效文献量相对较多,触发词方法的综合效果较优于模式匹配算法,但较多领域存在触发词算法的查全率较低的问题。
Event Extraction is a process which extracts a series of events which consists of time,place and character. First,the paper summarized the literature about event extraction,and gave the main method and model of event extraction. Meantime,it analyzed the results of these methods and explore the effect and application of these methods. The result was that,the research of Event Extraction was going on,which is focused on the internet,the We-Media,and Professional documents. The internet is about Financial information,conference information,unexpected events,personal resume. The We-Media includes Microblogs and We Chat. The professional documents are about Military,law. the algorithm have SMV,CRE,ME,Pattern Match,Cluster. the number of literature using pattern matching algorithm is more,the comprehensive effect of trigger method is better than that of pattern matching algorithm,but the recall ratio of trigger method is low in areas.