avukadin/biasAnalysisWithChatGPT

This code accompanies the white paper "Is the CBC a Neutral Voice in Canadian Politics?".

The final results used in the paper are in the data/results.csv file.

Code Outline

Below is an explination of the main files. Note that mose code is written in python, except for get_chat_gpt_response.js which is javascript. At time of writting, OpenAI does not have a API for ChatGPT the most convenient workaround was to use a javascript library. There is also the option of using text-davinci-003 which produces very similar results to ChatGPT.

1. get_article_links.py will grab article links from Google News RSS feed for articles published on cbc.ca/news and save them to data/links folder
2. get_articles.py will scrape cbc.ca/news site for the full articles and save them to data/articles.csv
3. extract_snippets.py will extract the article excerpts used for feeding into ChatGPT
4. classify_articles.py will query text-davinci-003 for classifying articles and save results to data/results.csv. To query ChatGPT instead, use node to run get_chat_gpt_response.js, this is not 100% complete as it will nto save to csv, instead print responses to console. This will be updated once an official API is released for ChatGPT.

White Paper- Is the CBC a neutral voice in Canadian politics? - Alex Vukadinovic.pdf