All three team members were interested in working on the misinformation problem as it relates to Coronavirus. The particular project was raised as a project idea from the "Pandemic Response Hackathon | Project Ideas" spreadsheet.
What it does
Scans online conversations to discover ‘how-to’ and guidance-based articles related to COVID-19 to help public health prioritize and target official campaigns, notices, clarifications to correct the disinformation and misnomers.
How we built it
We leveraged keywords from various public health articles such as pandemic response plans including Health Care Systems Sustainment, Surveillance, Vaccination, Community Containment, Laboratory Support, Telehealth, At-Home Testing Kits, Data Entry, Medication Distribution, Clusters, Fatalities, Protection, Workforce Infrastructure, Anti-Viral Therapy, Rationing, Ethical Response, Incubation Period, Seasonal Variation, Symptom Development, Transmission, Illness Rates, Absenteeism, Virulence, and Infectivity.
Using these keywords, we leveraged the PeakMetrics API to aggregate data from news and social media to spot articles related to guidance based articles. We automatically cluster the news articles or social posts that return around these keywords and surface the most popular results to the top based on the amount of social activity (ie Facebook shares, retweets).
Using this data, we hope that public health authorities can create comprehensive campaigns to combat bad information.
The site updates in real-time and functions with live data.
Challenges we ran into
Training data models around clustering. We tried multiple clustering algorithms to eventually end up using keywords as the base.
Accomplishments that we're proud of
Getting to an MVP that is actionable, usable, and updates in real-time!
What we learned
There is A LOT of misinformation out there and especially questionable content on YouTube.
What's next for COVID-19 "How To" Tracker
Continuing to have conversations with public health authorities and continue to build the solution for anyone interested.
Try It out
node.js, python, react