Analyzing COVID-Related Social Discourse on Twitter using Emotion, Sentiment, Political Bias, Stance, Veracity and Conspiracy Theories
Online misinformation has become a major concern in recent years, and it has been further emphasized during the COVID-19 pandemic. Social media platforms, such as Twitter, can be serious vectors of misinformation online. In order to better understand the spread of these fake-news, lies, deceptions, and rumours, we analyze the correlations between the following textual features in tweets: emotion, sentiment, political bias, stance, veracity and conspiracy theories. We train several transformer-based classifiers from multiple datasets to detect these textual features and identify potential correlations using conditional distributions of the labels. Our results show that the online discourse regarding some topics, such as COVID-19 regulations or conspiracy theories is highly controversial and reflects the actual U.S. political landscape.
This repository contains notebooks to train the three different models (sentiment, emotion, political bias) and to perform some data analysis.
- ./src/Data-Stats.ipynb contains some code to perform data analysis on the datasets, and to split them into stratified folds. It also convert the original datasets into normalized csv files with tweet content and label.
- ./src/Train-Inference.ipynb contains some code to train models on the datasets, to use them on other datasets (inference), and to visualise the results into heatmaps. The Load Data cells specify which dataset will be loaded.
Datasets should be copied to a new folder named ./data at the root of the repository.
Please cite this work as below:
@inproceedings{Peskine_BeyondFacts_2023,
author = {Peskine, Youri and Troncy, Raphael and Papotti, Paolo},
title = {{Analyzing COVID-Related Social Discourse on Twitter using Emotion, Sentiment, Political Bias, Stance, Veracity and Conspiracy Theories}},
proceedings = {3rd International Workshop on Knowledge Graphs for Online Discourse Analysis (BeyondFacts)},
doi = {10.1145/3543873.3587622},
year = {2023}
}
The heatmaps obtained with the different models can be seen on the following figures.
Sentiment:
Emotion:
Political bias:
