摘要
在线媒体拥有海量的非结构化数据,包含大量的突发话题和普通话题.传统的话题模型在不加入先验信息的情况下,无法有效区分普通话题和突发话题.本文的研究提出基于高效用模式和话题模型的突发话题发现(high utility bursty topic model,HU-BTM)模型.该模型使用高效用模式挖掘找出文本数据中的突发词组,使用基于普通Polya坛子模型的Gibbs抽样方法,将突发词组与突发词引入话题模型,实现突发话题的自动识别.实验结果表明与现有的主要突发话题发现方法比较,HU-BTM模型在准确率和召回率指标上优于对比算法.
Online media carry a huge amount of unstructured data,including a large number of common topics.Without prior information,the traditional topic model cannot effectively distinguish ordinary topics and emergent topics.In this paper,a high utility bursty topic model(HU-BTM)based on high utility pattern and topic model is proposed.In HU-BTM,bursty phrases and bursty words are found through high utility mining,and Gibbs sampling method based on general Polya urn model is used to introduce bursty phrases and bursty words into topic model to realize automatic discovery of bursty topic.The experimental results show that compared with the existing mainstream methods for discovering Bursty topics,the HU-BTM has better performances of precision and recall.
作者
闫志华
唐锡晋
YAN Zhihua;TANG Xijin(Academy of Mathematics and Systems Science,Chinese Academy of Sciences,Beijing 100190,China;University of Chinese Academy of Sciences,Beijing 100049,China)
出处
《系统工程理论与实践》
EI
CSSCI
CSCD
北大核心
2021年第5期1138-1149,共12页
Systems Engineering-Theory & Practice
基金
国家重点研发基金(2016YFB1000902)
国家自然科学基金(71731002,71971190)。
关键词
高效用模式
突发话题发现
话题模型
在线媒体
high-utility pattern
bursty topic discovery
topic model
online media
