partem.tech

We are the world’s first real time news source bias checking and sharing platform

Comment

Inspiration

The 2016 election came around just when we, young adults, were just starting to pay attention to politics. We had to face fake news, Russian hacker allegations, and email scandals. Trustworthy news was a high commodity. Not knowing what sources to trust, but having them thrown at us constantly, was a hassle. So we decided to solve this issue through analysis of news articles for anyone to verify what articles are safe from bias before you share these articles on social media or make an opinion.

What it does

Partem.tech’s platform utilizes powerful natural language processing algorithms to analyze intelligently scraped text from news sources and determine its bias (left, right, or center).

How I built it

We used boiler-pipe API to scrape news pages, google cloud platform for machine learning model training and web hosting, Xamrin forms and Aspnet for mobile application development for IOS, and python for data engineering/cleaning/science.

Challenges I ran into

One of our biggest challenges was to find a good quality-labeled dataset. Although we could not find any credible structured data sources, we used scraped tweets by Democrats and Republicans and manually labeled over 100 news articles scraped using boiler-pipe. The entire process was really cumbersome which led us to narrow the extent of our features and ambition. Extensive training time and its unpredictability was another issue.

Accomplishments that I'm proud of

We finally created a finished working prototype

What I learned

Our project idea turned out to be very ambitious for a 36-hour long hackathon which helped us push our intellectual boundaries in the fields we are interested in. We also learned learning skills on the fly!

What's next for partem.tech

We aspire to make partem.tech more accurate using a well-labeled dataset and more sophisticated machine learning tools. Apart from this, we see our app as a database of news articles, which helps our model learn over time and potentially help research institutes conduct better research on data mining and politics for improved political outcomes. We also see this as a useful product for journalists to check if there writing is biased or not and to what degree.

Built With

Share this project:

Updates