Please take a look at our data story.
All the analysis can be found in the repository.
The divide between Republicans and Democrats has long been a source of political tension in the United States. But in the age of modern media, platforms like YouTube have the power to shape and influence public opinion. With its big and diverse audience, YouTube has become a popular outlet for political soapboxing.
We set out to examine the political landscape of YouTube using the YouNiverse dataset, which includes metadata on 136,470 channels, 72,924,794 videos, and around 8.6 billion comments. Our goal was to understand how users are distributed among right- and left-winged major news channels on YouTube, whether the platform creates filter bubbles for users with specific political biases [1] [2] [3], and how certain YouTube trends relate to real-world events.
- Can we find evidence for the filter bubble phenomenon on the political YouTube?
- Does the political YouTube mirror real life politics?
- US news channels labelled with political bias from AllSides
- A collection of Donald Trump's approval rating polls gathered by FiveThirtyEight
- Data Preparation
- General Data Exploration
- Network Analysis
- Lexical categories
- Title sentiment for different keywords in video
- Filter bubbles (aggregate scores)
- First order connections
- LDA video titles
- Sentiment analysis pipeline
- scipy, sklearn
- huggingface transformers (https://huggingface.co/cardiffnlp/twitter-roberta-base-sentiment)
- spacy
- pyLDAvis
- gensim
- bokeh
- googleapi (youtube)
- Andrei:
- Filter Bubbles, Comment Data
- Data story
- Angelo
- LDA
- Data Story
- Juhani
- Sentiment Pipeline
- Time series
- Marc
- Read in Data & Pre-processing & Exploration
- Graph analysis and Visualization
- Sentiment Analysis Video Titles