Often times, news outlets with a political stance can be misguiding people with a one-sided story. We aim to tackle this issue by making it easy for people to hear alternative voices so we can prevent political polarization and biased views.
What it does
It allows users to paste an article URL into our web app called "unbiased?". Then, unbiased? will use a machine learning library to provide a visual diagram on whether the article is tilted towards left-wing or right-wing. Also, our web app will be provide an article that looks at the same topic from an alternative point of view.
How we built it
We trained a model using TensorFlow to help determine the political bias of a news article. On top of that, we used Azure's Bing News Search API to help us retrieve articles that are on the other side of the political spectrum. We used react.JS for the front-end as well as Figma to help design the site.
Challenges we ran into
We were having trouble training our AI to identify articles' political stance correctly. With the time constraint, we find it hard to allow our AI to cycle through enough data to allow it to be able to make accurate decisions. Nevertheless, after our hard work, we have managed to get to work with ~80% accuracy rate.
Given the short timeframe, it was difficult to get the necessary data to extensively train the model. The text scraped from the articles also need a lot of cleaning, so there is a lot of noise in the training data. Designing and training the neural network also took valuable time to train, time that could have been used to improve the front end.
Implementing the back end with the front end was difficult, as none of us had prior experience with TensorFlow and it's related challenges in creating good models as well as generating something that the front end could understand.
Accomplishments that we're proud of
We are proud of being here and creating something that can potentially change the world. Even though the overall process is very hard, we learned a lot today from our very helpful mentors and organizers. We believe in the future, we can take what we learned today and use it to create something even bigger.
What we learned
We learned how to deploy web-apps. Also, we learned how to effectively use TensorFlow and web scraping APIs like Diffbot to make our web-app function
What's next for Unbiased?
We will be adding more features as well as training our AI rigorously to allow it to be even more accurate and useful