Projects
Purple
View
Software Engineer
[04/2020 - 09/2020]
Newsletter scraping topics from media sources with different political perspectives and extracting their similarities. Backed by Stanford COVID-19 Response Lab.
I was at my parent's house, finishing my senior year of college in the midst of the Covid-19 pandemic. In between remote classes I couldn't help scrolling infinitely through social media and news outlets. The politicalization of the pandemic made it even harder to decipher what was actually happening in the outside world.
Along with 11 other Stanford students we created the purple view with the intention of bringing some peace and clarity to our lives and others. The idea was to aggregate as much news as possible as often as possible and using NLP isolate the similarities and filter out discrepancies. The final result was a free newsletter delivered bi-weekly to subscribers. The project gained backing by the Stanford COVID-19 Response Lab.
​
Development
- Sellenium datascraper collected the latest events from an assortment of news outlets.
​
- In order to distinguish each news source's political alignment we used the News Rating Data API by Ad Fontes Media.
​
- In-house NLP model isolated the similarities and presented them as bullet points. We measured the reliability of our generated "news article" by the metrics explored by Matthias Kohring and Jörg Matthes in their paper Trust in News Media: Development and Validation of a Multidimensional Scale.
​
- Subscribers received a bi-weekly newsletter broken down by topic and separated by political alignment (as seen above).