Social Status and Novelty Drove the Spread of Online Information During the Early Stages of COVID-19

This repository was created to host the code used for the research paper Social Status and Novelty Drove the Spread of Online Information During the Early Stages of COVID-19. We analyzed more than eight million randomly selected Twitter posts from the first wave of the pandemic to study the role of the author's social status (Health Expert or Influencer) and the informational novelty of the tweet in the diffusion of several key types of information. We provide the code and sample data to execute the given code for all of the main components of this study; novelty of information, social dimensions of conversations, and the regression framework used to examime the diffusion of information.

The repository contains 3 notebooks

• quantifying_novelty.ipynb: Contains the algorithm that was used to calculate the tweet-level novelty of information.
• social_dimensions_classification.ipynb: Contains the process of training machine learning classifiers to predict whether a tweet belongs to a specific category.
• regressions.ipynb: Contains the code for all the regressions that were mentioned in the paper.

The repository contains 5 datasets

• annotated_tweets.csv: The results of the manual annotations (status_id and social dimensions)
• annotated_tweets_data.csv: The tweet text collected from the Twitter API for the annotated tweets in annotated_tweets.csv
• dataset_for_regressions_sample.csv: A sample of the final dataset that was used for the regressions. Due to the size of the full dataset, we couldn't include the whole dataset in GitHub. The full dataset can be found at: https://www.dropbox.com/s/ai77p16i19uhzri/final_dataset.csv?dl=0
• raw_tweets_sample.csv: Random sample of tweets. Used in quantifying_novelty.ipynb. Due to the size of the full dataset, we couldn't include the whole dataset in GitHub. The full dataset can be found at:
• tweets_to_predict.csv: Sample of tweets to apply the machine learning models build in social_dimensions_classification.ipynb