We were inspired by Google's natural language processing capabilities, as well as the growing need for tools to help facilitate online political discussion.
What it does
Sensis takes articles that you visits and a scans it for emotional sentiment and political biases. It then returns the overall rating in bias to you in the form of a rating from 1-100, and then highlights parts of the article that helped sensis come to its conclusion.
How we built it
Sensis starts with a flask server to get and parse page html for article text, then sends that text to google's natural language processing api to get a sentiment analysis. An in house machine learning algorithm running on a separate server uses a set of training data to determine political bias in any given article. A chrome extension then injects data back into the page to display highlighting, and also renders an overall bias rating as well.
Challenges we ran into
Hooking together all of the components was particularly difficult. We didn't expect that we'd need both a node and a flask server, but it ended up being a fairly complicated data path.
Accomplishments that we're proud of
The in house machine learning algorithm as well as the google chrome plugin were some of our biggest achievements. We're proud to have learned to so much and accomplished as much as we did.
What we learned
We learned a lot of about networking, data analysis and found a plethora of unique apis along the way.
What's next for Sensis
Sensis's in house political bias training algorithm is far from perfect, but with a larger set of data and some training, it holds a pretty killer amount of potential for the future of political analysis.