Ethnic Bias in COVID-19 Related Articles

This project leverages word embeddings to quantify the ethnic biases in COVID-19 related articles.

Goals

This exploration includes investigating the associated bias for Asians in comparison to White ethnicities. This analysis is broken into the following sections:

1. Association with COVID-19 terms
2. Assocation with Hate Crimes
3. Association with Outsider Adjectives
4. Sentiment of Most Associated Adjectives

Data

The data was collected using AYLIEN's COVID News dataset. Data is presented in JSONL file format with one line per story object. To access the origina AYLIEN dataset, you can sign up to download it for free here: https://aylien.com/blog/free-coronavirus-news-dataset

Word lists were collected through a compilation of news articles and papers for COVID-19 and Hate Crime terms. Outsider adjectives and the larger list of general adjectives were used from the SI of Word Embeddings Quantify 100 Years of Gender and Ethnic Stereotypes.

Analysis

To re-run all of the analyses:

• Download the word lists from the data section
• Download the embedding models from the models folder
• Run the code in embeddings_ethnicCategories.ipynb, embeddings_wordCategories.ipynb and embeddings_largeAdjectives.ipynb to generate all of the analyses and plots

These notebooks are categorized into the three computational methods, first analyzing embedding bias with group vectors representing categories, secondly with group vectors representing categories and then lastly the sentiment analysis with a larger list of adjectives