We had a wonderful time exhibiting at APHA in Denver a few weeks ago. There is no replacement for being in person, as I mentioned in an earlier blog post. The kind of discussions and encounters we were able to have while discussing our work was extremely beneficial. There’s been something lost in the translation when it comes to things going online, but that’s a different topic.
AND for those who couldn’t be there, AND for those who didn’t attend online, Twitter might occasionally be a helpful tool in keeping up with the trends and themes.
A Twitter bot was created to monitor APHA related tweets during the conference. So we went through all of the tweets that used the hashtag #APHA2021 to get an idea of what people were talking about at least on Twitter during the conversation.
This blog post will highlight some of the most frequently used words (topics) throughout the course of this event. Some keywords are more specific than others, but there’s a nice range that provides insight into what people were into.
The first graph depicts the frequencies of various words and phrases that appeared across all tweets. It’s no surprise to see COVID-19 at the top, given its prominence in previous years. We were particularly interested to observe “carceral system” climb up, which is a phrase I’m not personally acquainted with. A carceral system pertains to the aspects of the prison system that extend beyond the gates of the prison and into the community.
Of all the forms of analysis used in computational linguistics, network analysis is one of the most popular. These typically perform well at extracting high-level themes, but in this instance, it’s a lot of nonsense. Looking below we can see some themes, but also a lot of the incidental information. Moving along…
To get a deeper understanding of the tweeter’s emotions, I used sentiment analysis to extract the emotional content of their tweets. Because the NRC lexicon separates positive and negative words into distinct emotions, I utilized it.
Here, we can see that Trust was the most frequently seen emotion, followed by anticipation. So, people were excited about their APHA experiences.
We can use topic modeling to find some more nuanced trends. These topics were chosen using an algorithm (Latent Dirichlet allocation) that automatically selects important aspects of Twitter conversations based on how frequently they appear together in tweets. The top clusters are shown in the graph below, with each cluster’s distinguishing terms listed on the side.
The first topic appears to be promotional, meaning tweets that talked about upcoming sessions. Other topics are more understandable. Other topics range from the mundane (social media) and compelling (COVID-19). But there’s also a lot of chatter on systematic racism in America!
Finally, I looked at who was tweeting the most. Far and away, the top tweeter was PublicHealthMap, which I thought was bot, but turns out that it’s a real set of people! They were busy.
And finally, the most liked tweet is embedded below.
I hope that this provides some insight into what people were discussing during APHA. If you were there, I hope that it’s helpful to see the content of what others were talking about.
Want more? Follow us on Twitter @pubtrawlr and let’s continue the conversation