[论文解读] Global Sentiment Analysis Of COVID-19 Tweets Over Time

Muvazima Mansoor, Kirthika Gurumurthy+2 more|arXiv (Cornell University)|Oct 27, 2020

Sentiment Analysis and Opinion Mining参考文献 8被引用 38

一句话总结

该论文使用 VADER 标签和 ML 分类器（LSTM 和 ANN）分析 2020 年 1 月至 2020 年 6 月的全球 COVID-19 推文情感，并在案例数据的背景下考察 Work From Home 和 Online Learning 的情感。

ABSTRACT

The Coronavirus pandemic has affected the normal course of life. People around the world have taken to social media to express their opinions and general emotions regarding this phenomenon that has taken over the world by storm. The social networking site, Twitter showed an unprecedented increase in tweets related to the novel Coronavirus in a very short span of time. This paper presents the global sentiment analysis of tweets related to Coronavirus and how the sentiment of people in different countries has changed over time. Furthermore, to determine the impact of Coronavirus on daily aspects of life, tweets related to Work From Home (WFH) and Online Learning were scraped and the change in sentiment over time was observed. In addition, various Machine Learning models such as Long Short Term Memory (LSTM) and Artificial Neural Networks (ANN) were implemented for sentiment classification and their accuracies were determined. Exploratory data analysis was also performed for a dataset providing information about the number of confirmed cases on a per-day basis in a few of the worst-hit countries to provide a comparison between the change in sentiment with the change in cases since the start of this pandemic till June 2020.

研究动机与目标

• Assess global public sentiment toward COVID-19 using Twitter data from Jan 2020 to Jun 2020.
• Evaluate sentiment trends for Work From Home and Online Learning during lockdown periods.
• Compare sentiment dynamics with daily COVID-19 case trends in selected worst-hit countries.
• Develop and compare machine learning models (LSTM and ANN) for tweet sentiment classification.

提出的方法

• Create three tweet datasets: coronavirus, online learning, and work-from-home from Jan 2020 to Jun 2020.
• Label sentiment using VADER and perform binary (positive/negative) sentiment analysis and emotion detection (fear, trust) via NRC EmoLex.
• Train and evaluate LSTM and ANN classifiers on labeled data for sentiment classification; report accuracy.
• Perform exploratory data analysis and visualize sentiment and emotion trends globally and for worst-hit countries.
• Correlate sentiment/emotion trends with daily COVID-19 case data from Kaggle datasets.

实验结果

研究问题

• RQ1 What are the global and country-level sentiment trends toward COVID-19 from Jan 2020 to Jun 2020?
• RQ2 How do sentiments toward Work From Home and Online Learning evolve during lockdown periods?
• RQ3 How do fear and trust emotions vary over time and across countries, and how do they relate to case trends?
• RQ4 How do LSTM and ANN compare in accuracy for classifying sentiment in COVID-19 related tweets?

主要发现

• LSTM achieved 84.5% accuracy on coronavirus tweets; ANN achieved 76% accuracy.
• Positive vs negative sentiment gaps were largest in Feb–Mar, narrowing in later months.
• Fear dominated the emotion scores, with trust also present; countries varied in emotion profiles (e.g., higher trust in some nations).
• WFH sentiment remained more positive than negative overall; Online Learning tended to be positive but with notable negatives.
• Case trends showed exponential growth after March, with sentiment/emotion patterns reflecting lockdown policies and case dynamics.
• Three datasets (coronavirus, WFH, Online Learning) enabled comparative visualization against global case data.