The Other Side

A screenshot of the extension in action
The LDA model we trained to cluster news articles by topic
Another demo!

Inspiration

If we've learned anything from the 2016 US Presidential Election, it's that we often don't fully understand the other side's opinions. Here's an easy way to better inform yourself!

What it does

The Other Side analyzes the content of the page you're currently browsing and scours the Internet for news articles about the same topic but written from the opposing political perspective. Snippets of these articles are presented in a pop-up.

How we built it

The back-end server is Python, and the machine learning is handled mainly by Microsoft Azure ML Studio and sklearn. The corpus of data we used was text data from manually labelled well-known websites.

Challenges we ran into

Firstly, it was extremely time-consuming and labor-intensive to gather the necessary labelled data for predicting political ideology. We scraped numerous news websites and learned how to use common web mining libraries such as Beautiful Soup. Furthermore, it is incredibly difficult to accurately predict political ideology from a small snippet of text, so acquiring an effective model required an extensive amount of cross-validation and experimentation. We used both Python's sklearn and Microsoft Azure ML Studio, both of which were new tools to us.

Accomplishments that we're proud of

We're really proud and excited that we managed to make our ML models work! It was a really interesting and important challenge, and we hope to be able to apply these skills further in the future.