An interactive visualisation of racism on Social Media

Interactive World Map
Start screen
Storyline 1
Storyline 2
Data visualisation
Description of major events

Inspiration

We've had several friends who've been the victim of racist attacks on social media in the past. Keeping this in mind, we wanted to analyse the trend of racist content on several social media platforms on the internet and link any peaks in racist content to major world events which may have influenced them. Furthermore, we wanted to make a project as interactive and accessible as possible.

What it does

We built an interactive storytelling styled educative platform, in which we present the data we scraped from the internet, after applying an LSTM machine learning model on it to classify particular texts in relation to its potential racial biases. The data was analysed on a regional basis, and was taken over a period of around 10 years. We then identified correlations between major social world events (for example the death of George Floyd in 2020) and any peaks in our data.

How we built it

For scraping the data we wrote a javascript script and made api calls to several social media platforms. The LSTM model was built using tensorflow in python with a training set from Kaggle. The graphs were plotted using matplotlib in python. The interactive components of our project were built using C# scripts in Unity.

Challenges we ran into

Being rate limited by the twitter API
Not having a very good training dataset
Limited amount of data in certain countries
Our model only works for texts in English (a lot of racist content on social media is in other languages)

Accomplishments that we're proud of

We developed an engaging interface for presenting our result
As per our knowledge, this seems to be the first long term analysis of social media content analysing racism separated by countries.

What we learned

How to scrape data from the internet
Building machine learning models
Building interactive world maps on Unity

What's next for Visualising racism on Social Media Platforms over time

Extending the project to more countries
Extending the project to other social issues (income inequality, sexism and so on)
Improving the ML model that we built by providing more data
Include present day/live tweets identify any upwards/downwards trends in racist content and similarly check if there happens to be a correlation between present day events