Orlando Shooting: What did Americans say on Twitter? | Initium Lab | Exploratory Arm of Initium Media

Orlando Shooting: What did Americans say on Twitter?

“Luis Vielma worked on the Harry Potter ride at Universal. He was 22 years old. I can’t stop crying.” This is the message retweeted 439,142,153 times in 36 hours on Twitter, after Orlando shooting tragedy. They were mourning for Luis, one of the victims.

At about 2 o’clock June 12, armed with an assault rifle and a pistol, Omar Mateen slaughtered 49 people and wounded at least 53 in Pulse, a gay nightclub in Orlando. According to CNN, the gunman pledged allegiance to ISIS and this is “the deadliest mass shooting in the United States and the nation’s worst terror attack since 9/11”.

What did Americans say after the attack? We have turned to Twitter to find out the answer. We totally scraped 358,063 tweets in the United States, under hashtags “#OrlandoShooting”, “#Orlando”, “#OrlandoUnited”, “#TwoMenKissing” and “#LoveIsLove”, through Twitter’s public API within 36 hours and analyzed their social media expressions with tableau.

Instant news does not mean instant reaction

image1

In the chart above, we plotted the people’s Twitter behaviors in accordance with time series. The X axis represents the hours on June 12 and 13 (since the incident happened during the night, there were not enough related tweets returned in the first 10 hours in our sample test). The bar (right Y axis) represents the tweet count and the line represents the retweeted count (left Y axis). It’s clear to see that in the first 24 hours of our sample, people did not tweet or retweet a lot on Twitter. The peak came from 17:00, June 13, 36 hours after the incident’s occurence. Although the first tweet about the news came out instantly, people’s reactions may experience a delay.

“Uncle Sam can not stop crying”

image2

The gunshot incident attracted the whole country’ attention. The map shows the result of each state’s average sentiment score level. The bluer a polygon is, the more negative words in local tweets. Most states in America are shadowed in a “blue” emotion. Just like the tweet we mentioned at the very beginning – “Uncle Sam can not stop crying”.

But some states also showed their warm hugs. Iowa, Mississippi and Michigan showed strong positive emotions. “We stand together with Orlando tonight” – a tweet from Michigan said.

Love beats Hate

image3

Based on tens of thousands of tweets, we then draw a word cloud to extract the most popular words in Americans’ tweets. Besides “orlando” and “shoot”, which are the indicators of this incident, “love”, “people”, “victims” are the most popular words. We also find that a lot of people mentioned “hate”. But this time, “love”, together with “prayers”, “vigil” and other warmhearted words won the opinion trend.

A heartbreaking moment raised up a united country

We filtered out the most retweeted tweets (retweet more than 50000 times in 36 hours) in our sample and aggregated those retweets who had same contents. And then a content analysis about emotions was made as below. The Y axis is the number of records of each emotion, and colors of bars represent the total retweet count of emotions.

image4

“My heart is broken for Orlando”, ”A mother received chilling text messages from her son who was inside Club Pulse in Orlando during the shooting”, “Chilling”, “deeply saddened by the tragedy in Orlando”…After the gunshot, the most retweeted emotion on Twitter is sadness. The lost of 49 innocent lives made the country crying. People expressed their sorrow for the tragedy.

However, among the most retweeted tweets (unique text), more individuals tweeted encouragement or comfort emotions themselves, rather than just retweeting others’. “I stand with Orlando”, “Standing as one in The Castro, San Francisco”, “The Bond Bridge will remain lit in rainbow colors in solidarity with Orlando. Stopthehate”, etc. After mourning for victims, people from all over the country began to give courage and comfort to the crying land.

A tweet precisely described Americans’ feeling at that time: “Feeling so sick, terrified, and heartbroken. I don’t know what to say, but love one another, hug one another. We stand in solidarity with each other.” A heartbroken moment raised up a united country and love is the main theme.

Methodology

We used Twitter public API to scrape data by locations. We first selected each state’ s biggest city to represent that state. Then the tweet-created time was limited between 12:00:00, June 12, to 23:59:59, June 13. For each city (50 cities in total), our rooftop is 10000. No matter how many tweets the API returned in each city, they will be returned randomly by the created time in the period. Moreover, our searching keywords included: “#OrlandoShooting”, “#Orlando”, “#OrlandoUnited”, “#TwoMenKissing” and “#LoveIsLove”.

As for the sentiment score calculation, we used the lexicon-based method. The dictionary comes from Liu Bing, Hu Minqing and Cheng Junsheng (2005), Opinion Observer: Analyzing and Comparing Opinions on the Web. You can find the resource here. The higher the scores, the more positive words the tweet has; the lower the scores, the more negative words the tweet has.

In the word cloud processing, we removed all retweeted tweets and only kept unique tweet text. “Orlando” actually appeared 5,1535 times which is the most frequent word. The size of “Orlando” was adjusted on purpose to fit the word cloud.

In the last part of the analysis, we employed two independent coders to code the emotion, since the sentiment package in R is not reliable enough (only about 20%~40% accuracy compared to manually coding result). We selected the most retweeted tweets which had been retweeted more than 50,000 times in the limited period. And then filtered out duplicated texts.

The emotion coding scheme is based on UCSC dreamsearch and you can find it here. We also add several emotions in our test: disgust (words\phrases represent disgust, i.e. shame, satire), fear (words\phrases represent fear, i.e. chill, afraid), encouragement or comfort (words\phrases represent encouragement or comfort, i.e. pray, stand with, together) and NA (can not decide emotions). Our Cohen’s Kappa is 0.722 which achieved a substantial agreement in inter-coder reliability.