基于层次观点树的社交媒体多维度观点挖掘研究

Multi-dimension Public Opinion Mining of Social Media Based on the Hierarchical Viewpoint Tree

挖掘社交媒体舆论中隐含的观点有助于人们快速有效地了解社交媒体舆论,避免主观和随意地发表评论,传播错误信息进而引发恶性事件。目前,社交媒体观点挖掘主要从观点主题、倾向性或某方面内容等单个维度分析舆论,人们难以全面认识舆论并掌握这些观点内容之间的逻辑关联等多维度信息,且各子任务的相关性能还有待提高。为了更准确地了解且综合地分析不同维度的舆论信息,促进人们对社交媒体舆论的深入认知,本文提出了一种面向社交媒体短文本,体现各维度观点内容之间逻辑关系的层次观点树构建方法,并选取推特(Twitter)中有关羟基氯喹治疗COVID-19疾病的话题内容,进行层次观点树构建的实证研究。结果表明,本文提出的层次观点树构建方法能够提供多维度、易理解的社交媒体观点信息。

Mining of public viewpoints on social media can help people quickly and effectively understand them,avoiding subjective and casual comments and spreading wrong information,leading to malignant events.Currently,viewpoint mining on social media mainly analyzes public opinion from a single dimension such as the theme,tendency,or aspect content of viewpoints.It is difficult for people to fully understand public opinion and grasp multi-dimensional information such as the logical relationship between these viewpoints;therefore,the relevant performance of each subtask needs to be improved.To more accurately understand and comprehensively analyze public opinion information of different dimensions and promote people􀆳s in-depth understanding of public opinion on social media,this article proposes a construction method of the hierarchical viewpoint tree on social media which reflects the logical relationship between viewpoints in various dimensions,and selects the topic of hydroxychloroquine as the treatment of COVID-19 on Twitter to conduct an empirical study on this topic.The results show that the construction method of the proposed hierarchical viewpoint tree can provide multi-dimensional and understandable viewpoints on social media.