The COVID-19 pandmemic hit the United States in early spring, 2020. As common in times of crisis, millions of Americans took to Twitter to share and discuss information about COVID-19.

We analyzed tweets originating in the United States1 between March 1 and March 30 of 2020 to track how the conversation evolved over time. Scroll down to explore

1. The COVID-19-TweetIDs dataset was collected via the Twitter API over the course of the pandemic. The dataset comprises 1% of the total Twitter converstaion, randomly selected by the Twitter API.

How much did different states tweet about COVID-19 in March? Hover over a state to see how many tweets mentioned COVID-19 every day in March 2020.

Population: 39,512,223
The number of reported number of tweets represents ~1% of total Twitter traffic in each state.
How does COVID-19 feel? Twitter users expressed a wide range of emotions about COVID-19.

The interactive map below reflects daily average sentiment scores in USA for every day of the month. Evidently, the shelter in place orders that came into effect in the middle of the month correlated with a systematic shift in sentiment across the country, with users tweeting more words like grateful and beautiful towards the end of the month.

Select retweets below to find out what kind of content users retweeted.
Vader sentiment analysis tool was used to score each tweet on a sentiment scale from -1 to 1. Averages were comupter over all tweets per state by day.
What did people talk about? Conversations on Twitter are often politically charged. Political discourse on social media can be polarized and prone to misinformation, which may have harmful effects in the public sphere2. Who plays a role in facilitating these conversations?

Ranking the top most frequently occuring sequences of #hashtags used when discussing COVID-19, we find that select accounts tend to re-use polarized hashtags multiple times over a short period of time and flood the platform.
14 out of the top most tweeted hashtag pairs are politically charged.
But only 5 remain when duplicate posts by users are discounted.
2. Matt Kapko, Twitter's impact on 2016 presidential election is unmistakable,
Grinberg et al, 2019. Fake news on Twitter during the 2016 U.S. presidential election
Connecting #hashtags that often appear in tweets together and making use of a neat clustering alrothim, we can build a network graph of the #COVID-19 Twitter discourse. Exploring how the hashtags are connected to each other exposes the diverse threads that make up the main #COVID-19 converstaion.

Hover over #hashtags in the graph below to explore how they are connected to each other.
What were the most buzzing #COVID-19 threads on Twitter every day in March?

By tagging tweets with specific #hashtags, users participate in the emerging conversations aroung these threads. To further understand how each thread ties into the main conversation, we can look at the relative amount of attention users assign to it.
We filtered out the top 50 most discussed threads and ranked them based on the number of #hashtags tweeted to the topic each day. In the plot below, we evaluate the relative importance of the most popular #COVID-19 thread each day in March.

Hover over the thread labels to explore each one separately.
Design for this visualization was borrowed form #Election2016: US Presidential Candidate Twitter Buzz