Estimating time-series changes in social sentiment @Twitter in U.S. metropolises during the COVID-19 pandemic

This repository contains the input and output data for the research article entitled Estimating time-series changes in social sentiment @Twitter in U.S. metropolises during the COVID-19 pandemic by Ryuichi Saito and Shinichiro Haruyama. The article was published in the Journal of Computational Social Science in 2022.

The collection_data folder contains tweets collected using the Full-archive Search API of Twitter API v2 and the result folder consists of folders tweets sentiment classified by the BERT model and the GPT-3 model and TF-IDF results. The folder of collection data and estimation data consists of 1 to 3 folders, which numbers mean types of state-government orders that restricted citizens' activities.

Folder No.	Related Tweets Types
1	Stay-at-home
2	Restriction-on-gatherings
3	Travel-restrictions