Visualizing Trump Tweets

EPFL COM-480 - Data Visualization Project

Vincenzo Bazzucchi, Christian Sciuto, Lorenzo Tarantino

Screencast Process book Github

What are you going to see?

The new President of the United States Donald Trump uses Twitter as his main communication channel with all US citizens (and not only them). He tweets about every current topic and issue and his opinion can reach a large number of people, thanks to his political and financial position, but also to his speaking/writing abilities.

We are interested in studying and analyzing the format, contents and emotions contained (and sometimes hidden behind what he writes) in his tweets. Since these tweets reach almost everyone that has a Twitter account, it seemed legit to us to go deeper in the their meaning and expressions, trying to find patterns, different and common usages in some of the most used topics touched by Trump.

From the 33k and more tweets that Trump ever made since 2009, we extracted only the ones that refer to nine different topics:

  • Internal Politics
  • Obama
  • Hillary Clinton
  • Foreign Politics
  • China
  • Business
  • Shows
  • Interviews and Debates
  • Various

For every topic we studied the usage over the time, the most used words, and the number of retweets and favorites (Twitter keeps changes this into stars/hearts). For every tweet we tried to perform a simple sentiment analysis: this dataset offered by the National Research Council of Canada, allowed us to build an overview of the emotions hidden in the president's words.

How is the visualization structured?

After clicking the button start, two buttons will be showed: "topics" and "sentiment". Clicking on one of them will show bubbles referring to different topics/sentiments depending on what you previously chose. For instance, if you click on "topics", all the topics above listed will be displayed, one bubble for each of them. Every bubble has a different size, depending on how many tweets are contained in that particular topic.

Once you can see all the bubbles, you can just click on one and see all the statistics explained above. If you are curious to compare different topics or different sentiments, you can drop two of them in the left black pane and see the differences between the two compared objects.

OK! Now you are ready to start. Don't be shy and click the button!

President Trump
Data

Considerations

Tweets Usage

We can observe some interesting patterns in the tweets usage.
Since the start of his campaign, we can see a strong decrease of tweets concerning business (almost 0). Clearly, his first goal was politics and not business anymore. On the other hand, from 2015 Trump's interest in Foreign Politics increased. Following the same trend, we can observe that tweets concerning Internal Politics are mostly concentrated in the period of the Campaign and Presidency.
There are a lot of tweets regarding Obama in 2012 (when he was reelected), and some outliers during 2014.
There are two peaks for tv shows and they both are in between two years (December 2012 and January 2013 and the same for 2014/2015). We can link this behavior to the renewal by NBC of the new seasons of the show produced and presented by Donald Trump, Celebrity Apprentice.

Word Usage

We can see that the most used words in every cluster reflect the meaning of the topic. For instance, in business the words golf, hotel, ties, tower and shirts (all properties and merchandise of Trump) are used a lot, in interviews we can see that the most used words are interview, people, foxnews, job, foxandfriends, and so on for every other topic. In shows there are a lot of references about Trump's show Celebrity Apprentice, in China Obama often is quoted as well as the oil issue.
For what concerns sentiments, it is interesting to note that the most used word for every emotion is Trump, except for "joy", where the most used words are "fox", "enjoy", "true". The usage of "fox" leads us to state that when he speaks about his interviews he is generally happier than in other contexts, even if the emotions associated to interviews and debate don't completely reflect this observation. Other relevant observations are that Obama appears to be most quoted in negative emotions such as "anger", "fear" and "sadness". As we have seen in the sentiment analysis of topics, business and shows are the two ones where trust levels are the highest: we can see this behavior in the most used words for the trust emotion: "apprentice", "nbc", "celeb", etc.

Sentiments

We can see that generally Trump has more positive than negative tweets, the public opinion approval might be good for business and politics. We can find a more specific example of this behavior by looking at the emotions in Business, Shows and Interviews. Of course, when he tweets about his business, he has to be positive: we can see an high level of trust in his tweets. For what concerns shows, we can state very similar considerations: he produces and leads a tv show called 'Celebrity Apprentice' and he often tweets about it. He needs to transmit positive vibes to convince people that his show is worth watching. A very similar pattern for similar reason is found in the interviews and debate topic. Concerning politics we have different trends: there are high levels of anger and fear, with an high level of trust, in every topic concerning politics except for Obama, which is a little different. When it comes to politics he generally has slightly more negative tweets than positive. The interesting thing is that his level of trust is always the highest in every topic compared to the other emotions as we expected, except when he talks about Hillary Clinton or Barack Obama. About the Clinton topic, the highest emotions are anger, fear, disgust and sadness, in this order. Note that it s not something that we were not expecting, but the fact that the trust level is lower than the four other negative emotions it's very relevant. In the Obama topic, positive and negative emotions are almost at the same level, even if sadness is the highest emotion we can find. Trust comes after sadness: he is more confident when he talks about Obama than Hillary Clinton. Note that the levels of joy are often low compared to the other emotions, except for shows and business.

Retweets and Favorite count

We can clearly see that the number of retweets and favorites are much higher when they are about Hillary Clinton. These tweets are from 2016 and 2017, so as we know, during the election campaign Trump's tweets reached a very high popularity.
Except for Hillary, other topics have similar trends of retweets and favorite counts. However, we can note that Internal Politics has the highest values among topics (that is because it also contains tweets concerning Hillary) and also that people are generally more interested in Foreign Politics and Interviews-Debates.
Looking at the interest of the public and at the sentiment categories, the most "loved" tweets are the ones where Fear is predominant. The next three most retweeted and liked sentiments are Anger, Disgust and Anticipation. It's interesting to see how tweets with a negative connotation are generally more followed by Trump's public than the ones with a positive connotation.

Thanks for your attention!