What it does
Our site is a fully-functional, message-board style social media feed capable of receiving large amounts of text-based, user-generated content. We use an assortment of techniques to convert complicated unstructured data into algorithm-friendly formats.
Basically, our site gives users the opportunity to notice varying degrees of bias in news sources, especially surrounding controversial topics.
How we built it
We started with a collection of natural language analysis papers and articles. We then chose techniques that produced the highest accuracy ratings in the shortest amount fo time. Kyle normalized and error checked a host of datasets to build a blacklist database with python. While Kyle worked with the data checking, Cyrus began on the API endpoints and webpage techniques. Cyrus worked closely with George to ensure clear communication between the Vue frontend, and consistent and accessible data propagation.
Challenges we ran into
We had plenty of ideas for rating articles, but many of them produced substandard results. We ended up scrapping some of our creative ideas to pursue the more effective ideas.
Accomplishments that we're proud of
What we learned
Pretty much everyone worked on something they had little to no experience with. From SPA JavaScript frameworks, dataset normalization, and web scraping, to training ML classifiers, each of us pushed the boundaries of what we knew and stepped outside of our comfort zones.