Inspiration
Our inspiration for developing the Cov-Misinfo Tracker was twofold: the first being the lack of resources related to determining root causes of health-related misinformation that we see online, which is particularly evident with COVID-19, and the second was our own drive to educate people about the pandemic. Fake news and other forms of misinformation can be found everywhere on social media, but how much of it is posted in earnestness? Do the people posting (and re-posting) this misinformation do so innocently believing it to be true, or are they responsible for its creation? Are they trolling, or are they woefully misinformed? If they are misinformed, how could we best help to combat this misinformation?
Google Cloud's Natural Language Processing API, in concert with Twint, a twitter scraping tool, and Firebase, a webapp development platform, gave us the opportunity to find out the answers to these questions ourselves.
What it does
The Cov-Misinfo Tracker identifies tweets that potentially carry misinformation about COVID-19 by keying in on the queries associated with it, as well as analyzes the tweets to determine key words typically used in non-factual COVID-19 statements and categorizes them based on how often they occur in non-factual statements. Using Twint, the app searches for tweets containing popular covid conspiracy theory related phrases, such as 'plandemic' and 'masks off america', we were able to find a seemingly endless supply of covid misinformation. Firebase's Firestore allowed us to quickly and easily store these tweets, and their associated metadata, for later processing and review. The Google Cloud NLP API gave us the ability to establish the veracity of the tweets we scraped, as well as identify which words showed up most often in genuine fake covid news
How our team built it
We used a Python structure and various other API's to make this work. Specifically, we used the Twint tool to query and extract specific tweets, the Google Cloud NLP API to analyze and extract key phrases that correspond to misinformation, and the Google Cloud Firebase API to store the analyzed information to later be displayed to the user.
Challenges we ran into
The two greatest challenges encountered as a team during the hacking phase were issues with API initialization and difficulties in integrating the disparate parts. Individually, we each encountered difficulties with environment setup for the APIs we had chosen to work with.
Accomplishments that we're proud of
We're happy to say that we were able to get our individual tools working, and were able to integrate them with relative success. While it's not an automatic transition between the various API's and tools, it's pretty close! The Twint output and natural language processing work together smoothly and export to the Firebase, which would then be displayed back to the user.
What's next for the Cov-Misinfo Tracker
We think there is a definite need for an application like the Cov-Misinfo Tracker. Misinformation isn't just limited to the COVID-19 pandemic and when the next pandemic hits we'd like to have a functioning system in place to provide an ease of mind as well as deliver correct facts to people. Additionally other industries may benefit from using a dedicated misinformation system similar to ours. Going forward we plan to completely automate most functions of the Cov-Misinfo Tracker, complete the integration process with Google Cloud services, and expand the abilities of the webapp.
Check out the GITHUB and youtube demo!
Team members: Aziz Uddin, Adithya Shankar, Ryan Walsh
Log in or sign up for Devpost to join the conversation.