aandyrea/master-thesis

This repository contains code needed to replicate the master thesis study, "Refugees Welcome? A comparative sentiment analysis of tweets in Germany surrounding inflows of Syrians and Ukrainians", by Andrea Cass

Master-thesis

Overview

This file explains the steps to reproduce the research carried out for my Master Thesis, titled "Refugees Welcome? A comparative sentiment analysis of tweets in Germany surrounding inflows of Syrians and Ukrainians".

The project aims to collect and analyze immigration-related tweets published in Germany in English and German during two time frames:

1. The Syrian refugee inflow
2. The Ukrainian refugee inflow

All steps are documented in the Notebooks:

• 01_Data-Collection_limited_Syrian.ipynb
• 01_Data-Collection_limited_Ukrainian.ipynb
• 02_Pre-processing_limited_merged.ipynb
• 03_Sentiment-Analysis_limited_merged.ipynb
• 04_Data-Preparation_limited_merged.ipynb
• 05_Exploration-and-Visualization_limited_merged.ipynb
• 06_Regression_limited_merged.ipynb

Installation

Please refer to requirements.txt

Technologies

This project was created with:

• Python 3.9.13

Notebook 03_Sentiment-Analysis_merged.ipynb was carried out in:

• Google Colab

All other Notebooks were carried out in:

• JupyterLab 3.4.8

Notebooks

01_Data-Collection_limited_Syrian

The collection of tweets during the first time frame--the Syrian refugee inflow--is documented in this notebook. Both English- and German-language tweets are obtained.

01_Data-Collection_limited_Ukrainian

The collection of tweets during the second time frame--the Ukrainian refugee inflow--is documented in this notebook. Both English- and German-language tweets are obtained.

02_Pre-processing_limited_merged

In this notebook, all collected tweets are pre-processed for sentiment analysis.

03_Sentiment-Analysis_limited_merged

Sentiment analysis is carried out in this notebook. NOTE: This notebook was produced and run in Google Colab. Therefore, it is recommended you run it in Colab rather than Jupyter.

04_Data-Preparation_limited_merged

The data is prepared for exploration and eventual regression in this notebook.

05_Exploration-and-Visualization_limited_merged-negative-binary

Here, the data is explored and visualized.

06_Regression_limited_merged-negative-binary

The final step of the study is carried out in this notebook. A logistic regression is conducted to predict sentiment based on inflow (Syrian or Ukrainian) in addition to two control variables: (1) change in foreign share of the total population and (2) GDP volume growth.

07_Revisions

Several adjustments were made based on the suggestions from my supervisor.