基于要素图的舆情事件线索化方法研究

发布时间：2018-03-26 05:15

本文选题：舆情　切入点：事件线索化　出处：《情报杂志》2017年07期

【摘要】：[目的/意义]将事件按照内在关联组织成有序结构的过程称为事件线索化,事件线索化能够展现事件间的因果联系与演化过程,有助于理解舆情事件的起因与背景,预测事件的后续发展,为舆情事件的处理提供依据。现有事件线索化方法主要依据事件要素的共现特性,然而部分存在关联的事件之间共现特性并不明显,导致现有方法存在局限性。[方法/过程]利用百度百科中的条目及其中的关键词来扩充事件要素,建立了由关键词与超链接构成的要素图模型,并提出了基于该模型的舆情事件线索化方法。所提方法通过要素图差异性计算与密度聚类两个阶段将事件集划分为由相互关联事件组成的簇,从而实现事件线索化。[结果/结论]在舆情事件数据集上的实验表明,所提方法在共现特性不明显情况下能有效地实现事件线索化。
[Abstract]:[Objective] the significance / event according to the internal connection process organized into ordered structures called event clues, clues of events can show the causal relationship and evolution between the events, to help understand the cause and background of public opinion of the event, the subsequent development of predicted events, provide the basis for treatment and the co-occurrence of parts. The main characteristics of event based elements of existing event clues method, however some association between event co-occurrence characteristics is not obvious, resulting in limitations. Method / process using Baidu encyclopedia entries and the key word method to expand the existing event elements, established elements of graph model composed of keywords and links, and put forward the model of public opinion event clue method based on the proposed method by factor graph difference calculation and density clustering will be two stages of the event set is divided into interrelated. Part of clusters, so as to realize the clues. Results showed that the experimental event / Conclusion] in the event of public opinion on the data set, the proposed method can effectively realize the clues in the event of co-occurrence characteristics is not obvious under the circumstances.

【作者单位】：西安交通大学城市学院;西安交通大学电子与信息工程学院;西安交通大学图书馆;
【基金】：教育部人文社会科学研究青年基金项目“新闻事件的线索化方法研究”(编号:15YJCZH057) 陕西省社会科学基金项目“开放环境下知识的刻面组织方法研究”(编号:2016N005) 西安交通大学城市学院校级科研项目“大规模在线学习中的隐性知识挖掘方法研究”(编号:2016KZ01)
【分类号】：TP391.1

【参考文献】