Game of Thrones has finally wrapped up after 8 seasons. Naturally, watchers have been outspoken about their opinions of the series finale and the final season altogether.
Now that we’ve had a few days to recover, we wanted to analyze watcher reactions and see if we could add some structure to our understanding of these responses. And what better medium to analyze than Twitter?
Leveraging some helpful tutorials from Julia Silge, Michael Kearney and Jessica Peterka-Bonetta, we were able to quickly collect and analyze Twitter data using R.
This was primarily an exploratory exercise, so our first step was to get data, using the rtweet package and a Twitter account. Tweets were scraped with the #GameOfThronesFinale hashtag and sample code is included below.
library(rtweet)
q <- "GameOfThronesFinale"
## Stream for 30 minutes
streamtime <- 30 * 60
## Filename to save json data (backup)
filename <- "tweets_got.json"
tweets <- stream_tweets(q = q, timeout = streamtime, file_name = filename, verbose = TRUE, parse = FALSE)We ended up with approximately 80,000 tweets, after removing retweets, which includes reactions immediately before and after the final episode. We used this tutorial for emoji analysis as well as this code.
The following sections detail noticeable differences in tweets before and after the last episode.
Top 10 Emojis – Before
| description_before | emoji |
| red heart | ❤️ |
| crossed swords | ⚔️ |
| female sign | ♀️ |
| male sign | ♂️ |
| white medium star | ⭐ |
| white heavy check mark | ✅ |
| snowflake | ❄️ |
| heavy check mark | ✔️ |
| frowning face | ☹️ |
| victory hand | ✌️ |
Top 10 Emojis – After
| description_after | emoji |
| red heart | ❤️ |
| female sign | ♀️ |
| male sign | ♂️ |
| snowflake | ❄️ |
| crossed swords | ⚔️ |
| heart suit | ♥️ |
| wheelchair symbol | ♿ |
| frowning face | ☹️ |
| white medium star | ⭐ |
| white heavy check mark | ✅ |
Sentiment Analysis
Comparing sentiment of tweets before the finale to that of tweets after the finale, we see a noticeable shift towards negative sentiment. This shift most likely includes a combination of low satisfaction levels, which have been well documented across review sites, to general feelings of sadness at the conclusion of such a beloved series.
In order to understand more context around the negative and positive sentiment, we used the tidytext package to extract the most used negative and positive words before and after the series finale.
Most Used Words – Before
Not surprisingly, prior to the episode, the most common words and phrases reflected an anticipatory mood. This includes both negative (anxious) and the positive (ready, patiently) sentiment.
Most Used Words – After
After the show, there’s a noticeable increase in the frequency of negative words, specifically around feelings of disappointment.
