Ever felt bombarded by social media posts? In today's fast-paced climate, it feels difficult to gauge others' opinions on important topics such as the COVID vaccine. If only information could be condensed into an all-encompassing, user-friendly visualization tool that can help vaccine providers and marketing professionals determine appropriate geographic areas to focus their efforts on. We designed Voices of the Vaccine to address these concerns and help users develop a better understanding of the vast spectrum of global opinions on the vaccine.


Voices of the Vaccine employs deep learning techniques to classify the positivity levels of relevant tweets in real-time and displays the ratio of positive-to-negative tweets in a particular area using a color gradient. Users can interact with the map by navigating to different areas of the globe.

How We Built It

We used NLTK and Vader to classify tweets pulled from a public COVID Vaccine Tweets dataset in order to build a dataset to train our neural network on. We then utilized TensorFlow to develop an accurate recurrent neural network to classify tweets scraped from the Internet. Using Plotly, we created a map that uses location tags from the tweets to display color-coded circles that represent the overall opinion of an area.

Challenges Faced

None of us had worked with Python for data analysis or to create applications before, so it was daunting to attempt to digest deep learning algorithms, natural language processing, and sentiment analysis in a short timeframe. We became frustrated when large dataframes took unexpectedly long to process and train. It was also challenging to learn how to make API requests and use Dash/Flask to create web apps with Python. However, we were able to overcome our obstacles and create a functioning, visually appealing web app.


We are proud of creating and interactive and comprehensive map from a large dataset. We harnessed Python and were able to conquer a large dataset in a visually appealing fashion

Key Takeaways

We were pleasantly surprised to learn that complex subjects such as deep learning and neural networks can be handled in a straightforward manner by taking advantage of handy features present in Python packages.

What's next for Voices of the Vaccine

It would be fascinating to incorporate a feature that overlays a map displaying the percentage of the local population that has received the vaccine and compare it to the population's overall sentiment towards the vaccine.


  • Our model makes it extremely easy to identify geographic areas that are experiencing more negative sentiment about vaccination during a given period of time
  • Healthcare officials can refocus their efforts towards educating local citizens about the benefits of vaccination
  • Ordinary users can initiate and participate in discussions to encourage others to vaccinate in their local community and gain a better understanding of global sentiment towards the vaccine

Built With

Share this project: