面向新闻领域的蒙古文新事件检测方法的研究  

Research on new event detection methods for mongolian news

作　　者：王士杰 飞龙 高光来[1] WANG Shi-jie;FEI Long;GAO Guang-lai(Mongolian Unirversity,IT Academy,Inner Mongolia,Hohhot 010021,China)

机构地区：[1]内蒙古大学计算机学院,内蒙古呼和浩特市010021

出　　处：《青海师范大学学报（自然科学版）》2020年第1期31-37,共7页Journal of Qinghai Normal University(Natural Science Edition)

摘　　要：新事件检测的目标是从一个或多个新闻源中检测出报道一个新闻话题种子事件的第一个新闻.本文面向新闻领域,研究了蒙古文新事件检测的相关方法.提出结合新闻内容相似度与新闻要素相似度进行新事件检测的方法.在新闻内容表示上,针对新闻的特点以及不同类型新闻表达词汇对传统的TF-IDF方法进行了改进.其次,提取出新闻的主要要素,包括时间、地点、主体、客体和主客体动作,计算了两篇新闻文档之间新闻要素的相似度.最后将新闻内容相似度与新闻要素相似度进行融合后进行相似度比较.实验结果表明,本文改进方法明显,性能比传统新事件检测系统有显著提升.New event detection(New Event Detection)aims at detecting the first news from one or multiple streams of news stories.This paper is aimed at the field of journalism and studies the related methods of Mongolian new event detection.The paper proposes a method that combines the similarity of news content with the similarity of news elements to detect the new event.For the news content representation,according to the characteristics of the news and the different types of news vocabulary expressions,improve the traditional TF-IDF method.In addition,extract the main elements of the news,including time,place,subject,object,denoter,and calculate the similarity of news elements between the two news documents,and combine the similarity of news content with the similarity of news elements to compare the final similarity to new event detection.

关 键 词：新事件检测 蒙古文 新闻要素 相似度计算 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]