This repository contains the code and data used for sentiment analysis of Sinophobia on social media platforms during the COVID-19 pandemic. The project leverages large language models (LLMs) to analyze tweets and track the rise of Sinophobia.
The repository is organized into several directories and files, each containing code, data, and visualizations related to the sentiment analysis project:
-
Sentiment_labelling.ipynb/
This notebook contains code which preprocesses and filters the data for the sample dataset of Australia, and labels the sentiments to each tweets. -
BERT_model/
This directory contains the Jupyter notebook used for the BERT model in sentiment classification. It includes:BERT.ipynb: Jupyter notebook that handles training and fine-tuning the BERT model for sentiment analysis tasks.
-
Results/
This folder contains CSV files with the results of the sentiment analysis for the selected six countries. Including original tweet, text after preprocessing, and sentiment labels. The results files can also be accessed here: https://drive.google.com/drive/folders/1OPMvmUBEsq-L70KSfNuQlk3uteyzG4eN?usp=sharing -
Sample_OriginalDataset/
This directory contains the sample dataset of the six countries, each contains the first 100 rows of the dataset. -
Visualisation/
This directory contains Jupyter notebooks dedicated to generating visualizations for the analysis, such as polarity scores and n-grams:Polarity score.ipynb: Jupyter notebook for calculating and visualizing polarity scores over time.visualisation.ipynb: Jupyter notebook containing additional visualizations, such as n-grams and trends by country.
Lande, Janhavi and Chandra, Rohitash, Global COVID-19 X (Twitter) Dataset (July 01, 2024). Available at SSRN: https://ssrn.com/abstract=4895603 or http://dx.doi.org/10.2139/ssrn.4895603
Wang, C., & Chandra, R. (2024). A longitudinal sentiment analysis of Sinophobia during COVID-19 using large language models. arXiv preprint arXiv:2408.16942. https://arxiv.org/abs/2408.16942