We became inspired after realizing the complexity of bias in media and the implicant bias people have when writing articles. We began the project by recognizing four major issues with bias in media:

  • Media companies are motivated to create biased articles to attract attention
  • Individuals introduce implicit biases unintentionally in their writing.
  • People can waste time reading articles without knowing the biases contained within those articles beforehand.
  • No standard API despite recent advances in natural language processing and cloud computing.

What it does

To address the issues we identified, we began by creating a general-purpose API to identify the degree and sources of bias in long-form text input. We then tried to build several tools around the API to demonstrate what can be done, in hopes of encouraging others.

Sway helps identify bias in the media by providing a chrome extension. The extension provides a degree of bias and the cause of the bias (if any exists) for any website, all with the click of a button. Sway also offers a writing editor to identify bias in the text. Lastly, Sway offers a search interface to conduct research on unbiased resources. The search interface utilizes Google search and our bias detection to rank websites according to credibility.

How we built it

We decided to build two microservices, along with a Chrome extension. We built the Chrome extension in HTML/CSS, and Javascript. The web platforms were built in React and Python. Our API for performing the text analysis was written in Python.

Challenges we ran into

We had trouble finding training data to properly classify particles. As we overcame this issue the training data didn't give us the outputs we expected so we had to spend additional time fine-tuning our model. The biggest challenge we ran into was creating endpoints to properly connect the frontend and backend. The endpoints had to be rewritten as we had to narrow down the scope of some features and we decided to scale-up other features.

Accomplishments that we're proud of

We built out three different platforms in a day! Sway also works extremely well and we're very happy with the performance. We had planned to only build the chrome extension but picked up good momentum and were able to add the two other platforms over the weekend!

What's next for Sway

We want to market and develop our algorithm further to the point where we could commercially deploy it. One thing that's required to do is add significantly more training data! We would like to keep adding more features to our core product. We would like to provide more helpful insights to users, such as numerical scores across multiple categories of bias.

Share this project: