Discovering Bursty Events Based on Enhanced Bursty Term Detection
Weibo has become preferred media for people to expose events,express opinions and share experiences.Many real-world events are first revealed on Weibo.Bursty event detection based on microblog has become a research hotspot in the area of recent event detection.This paper proposes a bursty event detection method(Burst_NBT)based on enhanced bursty term detection which is composed of construction of meaningful string dictionary and calculation of bursty term score.To utilize the feature of hashtags in Weibo and features of titles mark in Chinese,Burst_NBT adopts meaningful strings between “#s and meaningful strings between quotation marks as a heuristic method.The enhanced bursty term detection method takes three factors into consideration,word frequency,associated users and number of comments.Based on this,a hotspot computational model for bursty events is further developed,which uses the indexes such as bursty term frequency,associated users and hotness of associated posts.Experiments on the Sina Weibo corpus prove that Burst_NBT exceeds the other four bursty event detection methods.
Meaningful string Bursty term detection Bursty event detection
Liyan Zhou Junping Du Wanqiu Cui Zhe Xue Chengcai Chen
Beijing Key Laboratory of Intelligent Telecommunication Software and Multimedia,School of Computer S Xiaoi Research,Shanghai Xiaoi Robot Technology Co.,Ltd.,Shanghai 201803,China
国际会议
江苏镇江
英文
656-663
2019-09-20（万方平台首次上网日期，不代表论文的发表时间）