[论文解读] CoronaVis: A Real-time COVID-19 Tweets Data Analyzer and Data Repository

Kabir My, Sanjay Madria|arXiv (Cornell University)|Jul 11, 2020

Misinformation and Its Impacts参考文献 28被引用 22

一句话总结

CoronaVis 是一个实时的 Twitter 数据分析器与存储库，用于处理基于美国的 COVID-19 相关推文，以可视化主题建模、主观性及人类情绪。它提供了一个清洗过的公开数据集，并支持动态分析，以研究疫情期间的社会心理与行为，助力危机管理与未来疫情准备。

ABSTRACT

Due to the nature of the data and public interaction, twitter is becoming more and more useful to understand and model various events. The goal of CoronaVis is to use tweets as the information shared by the people to visualize topic modeling, study subjectivity, and to model the human emotions during the COVID-19 pandemic. The main objective is to explore the psychology and behavior of the societies at large which can assist in managing the economic and social crisis during the ongoing pandemic as well as the after-effects of it. The novel coronavirus (COVID-19) pandemic forced people to stay at home to reduce the spread of the virus by maintaining social distancing. However, social media is keeping people connected both locally and globally. People are sharing information (e.g. personal opinions, some facts, news, status, etc.) on social media platforms which can be helpful to understand the various public behavior such as emotions, sentiments, and mobility during the ongoing pandemic. In this work, we develop a live application to observe the tweets on COVID-19 generated from the USA. In this paper, we have generated various data analytics over a period of time to study the changes in topics, subjectivity, and human emotions. We also share a cleaned and processed dataset named CoronaVis Twitter dataset (focused on the United States) available to the research community at this https URL. This will enable the community to find more useful insights and create different applications and models to fight with COVID-19 pandemic and future pandemics as well.

研究动机与目标

• 利用实时的 Twitter 数据，理解公众在 COVID-19 大流行期间的情感、情绪与行为模式。
• 开发一个实时应用程序，以可视化公众话语中随时间演变的主题、主观性与情绪趋势。
• 创建并共享一个经过清洗、处理的美国聚焦型 COVID-19 推文数据集，以供更广泛的研究使用。
• 通过分析社会对疫情相关事件的反应，支持公共卫生与危机管理。
• 通过提供可重复使用、高质量的数据集与分析框架，支持未来对大流行的研究所需。

提出的方法

• 系统摄取聚焦于美国基于 COVID-19 的推文实时流。
• 应用自然语言处理技术，从推文中提取并分类主题。
• 执行主观性与情感分析，以评估推文的情感基调与事实内容。
• 使用情绪检测模型，对推文中表达的人类情绪进行分类。
• 数据处理管道对推文进行清洗与结构化，以支持长期分析与公开共享。
• 一个实时网络应用程序可视化随时间推移的主题、主观性与情绪的时序趋势。

实验结果

研究问题

• RQ1 在美国社交媒体话语中，与 COVID-19 相关的公众话题如何随时间演变？
• RQ2 公众推文中主观性与情感的分布情况如何？
• RQ3 在关键疫情事件期间，恐惧、焦虑或希望等人类情绪如何变化？
• RQ4 能否从实时社交媒体数据中推断出公众行为与认知的模式？
• RQ5 经过筛选的美国 COVID-19 推文清洗数据集如何支持未来研究与模型开发？

主要发现

• 该系统成功捕捉了公众话语中与疫情相关主题的动态演变。
• 推文中主观性水平存在显著差异，表明公众对事实性与观点性内容的参与程度不一。
• 情绪趋势显示出可检测的波动，与重大公共卫生事件及政策变化相对应。
• CoronaVis 推文数据集已公开提供，支持可复现的研究与模型训练。
• 实时分析揭示了情绪基调与特定疫情相关事件之间的相关性。
• 该平台证明了在公共卫生危机期间，利用社交媒体进行大规模行为与心理监测的可行性。