It has been fifteen months since the resignation of former President Hosni Mubarak in February 2011. Since then there has been a continued sense of leaderlessness and overall instability throughout Egypt. Over the last two days1, Egyptians have taken to the ballot box in what has become an historic presidential election. Hours since the voting polls have closed, the ballots are still being counted as Egyptians wait with bated breath. This race has been hard to predict up to the very last hours—as though the elections for Egypt’s next president have been held up by the lack of polling.
This essay examines social media content leading up to the presidential elections in May 2012. It provides ten interactive graphs to illustrate public opinion expressed on Twitter. These graphs represent sentiment and semantic analyses of over two million tweets from seventeen hashtag feeds posted from 10 April 2012 through 24 May 2012. The following hashtags are in the study: #egyelections, #egypresiden, #egypt, as well as the Arabic hashtags شفيق# حمدين# ,انتخابات_الرئاسة# ,العباسيه# ,ابوالفتوح# ,خالد_على# , موسى# ,مرسي# ,#مصر.
The results of this case study were generated using swarm computing algorithms developed at R-Shief, Inc.—a lab that is harvesting “one of the largest repositories of Arabic-language tweets” since 2008.2 This technique of data exploration works from the bottom up, filtering through the noise and letting the data tell us what is there. Leveraging its massive database, R-Shief was able to develop language, sentiment, and semantic analytic tools in Arabic and English beginning in 2008 (adding Persian, French, German, Spanish in 2011, and soon to include Urdhu) and applied in real-time analyses on political mobilization across the Middle East and North Africa region.
The data visualizations generated seek to illustrate and improve our understanding of the sensibilities and cultural logic(s)3 that are being expressed by the people on Twitter. It is not to say philosophical underpinnings to the nature of a virtual world are new and revelatory; nor does this argument purport that what is being expressed online in the digital world is necessarily representative of what happens on the ground. In places like Egypt, where literacy rates only reach sixty-six percent4, analyses of Internet penetration hold less weight5.
However, elements of the virtual become actualized under unique, local, and temporal conditions that cannot be predicted. They only happen in the “now.” Approaching this logic from a visual arts lens, as Laura Marks does in her book, Enfoldment and Infinity: An Islamic Genealogy of New Media Art, she traces new media art along a unique historiography of Islamic thought, from the birth of the algorithm in ninth century Iraq, through fifteenth century Islamic mysticism and neoplatonism, or “beginnings of virtual reality.” One of the critical points Marks builds upon is a notion of events in time as unique and foldable, similar to Kant’s nineteenth century notion of the “sublime” event. For an event to be transformative, it relies on unpredictable conditions. In other words, the act of Bouazizi lighting himself on fire in Tunisia was as sublime as it was horrible.
For the purpose of this essay, the event under examination is the first round of Egypt’s historic presidential election to replace former President Hosni Mubarak.
Graphing and Reading the Presidential Elections
Despite the fact that the majority of tweets are in Arabic, the English hashtag #Egypt is used in over twice as many tweets as its native Arabic equivalent, #مصر, according to the table in Figure 1. From the first day sampled for this study, 10 April, the ex-prime minister Ahmed Shafiq has had the most daily mentions of all the candidates. Shafiq has steadily been mentioned double and triple the volume as his opponents.
Fig. 1. These are the 17 Twitter hashtag feeds examined in this digital case study.
In order to determine whether volumes of tweets about a candidate indicate support or criticism, the sentiment analysis in Figure 2 indicates degree of positive sentiment expressed. However, without careful semantic consideration, conducting sentiment analysis can result in inaccurate results and in the past has proven to fail to detect sarcasm in tweets6. In the interactive bubble chart below, a new method of sentiment analysis was conducted on the Arabic-language tweets7. What was calculated, but is not represented in the chart below, are percentages of neutral and negative tweets.
Readers of this article are highly encouraged to roll the cursor over the all of the figures in this article to interact with the various analyzed data charts. When you roll over the charts with your mouse, detailed information will appear. Each bubble represents tweets collected on hashtags with the candidates’ name in Arabic. The size of the bubble represents the average volume of tweets in comparison to the others, and its position along the x-axis represents percentage of positive sentiment. The further to the right the bubble falls, the more positive sentiment detected in those tweets. When you roll over each bubble, a window of statistics will appear including a short description about the candidate.
From the sentiment analysis below, one would say that Ahmed Shafiq and Mohamed Morsy will move on to the next round of elections—Shafiq because of the sheer volume of public interest (whether positive, negative, or neutral), and Morsy because he came out with 99.83% positive sentiment. According to this bubble chart, the only contender is Hamdeen Sabbahi.
[Run your mouse over the graph to display comparative values]
Fig. 2. Created by VJ Um Amel, 2012. Research tools by R-Shief, Inc.
Fig. 2. Created by VJ Um Amel, 2012. Research tools by R-Shief, Inc.
The interactive bar graph below, Figure 3, gives you a weekly comparison of the candidates. If you roll your cursor over the columns, windows with more detailed information appear. This graph compares the average volume of tweets on the top candidates to be the next President of Egypt over the last week of campaigning. One thing clearly stands out: Shafiq, Sabbahi, and Morsy all gained more Twitter “talk” over the two last weeks. Mohamed Morsy seems to have increased the most from week to week.
[Run your mouse over the graph to display comparative values]
Fig. 3. Created by VJ Um Amel, 2012. Research tools by R-Shief, Inc.
Figures 4, 5, and 6 are interactive line charts that graph tweets over time. If you roll your cursor over the columns, windows with more detailed information appear. Figure 4 is a comparison among the Arabic hashtags for each candidate candidates zoomed into the 48 hours of Election Day one and two.
In Figure 4, there are two peaks evident. The first one occurred on 23 May around 5:00 pm Cairo time when Shafiq was hit by shoes thrown at him as he cast his ballot. The second peak of activity occurred in the final hours when Shafiq appeared as a clear winner. Alongside Morsy, Sabbahi and Moussa continued to gain Twitter talk as the results started trickling in Thursday night.
[Run your mouse over the graph to display comparative values]
Fig. 4. Created by VJ Um Amel, 2012. Research tools by R-Shief, Inc.
Figures 4 and 5 graph the number of times each candidate is mentioned in #Egypt tweets from 1 May 2012 through 24 May 2012. I chose to use the #Egypt tag because it receives the largest number of tweets in comparison to the other hashtags in the study (as an important rule of thumb, larger data samples allow the analyses to be more precise).
In previous studies, R-Shief has recorded that over seventy-five percent of tweets with #Egypt tags are written in Arabic. That understanding led me to graph the candidates name as it is written in English separately from graphing their names in Arabic. Though these two graphs seem similar, please note the numbers on the y-axis of the graph indicate different scales.
As we look at candidates` Twitter popularity of a timeline, one day that stands out is 10 May, when Amr Moussa and Abdel Moneim Abou El-Fotouh clashed in national television debates broadcast live. The debates seemed to have hurt the former two leading candidates, because soon after Sabbahi began to trend on Twitter while Moussa`s numbers dropped. Among English speakers, Sabbahi gained the most interest. The question thus follows: did he win over Moussa supporters after the debate?
Another striking observation is that Shafiq and Morsy were rarely mentioned by name in English throughout the month of May; however, at present, they are in the lead.
[Run your mouse over the graph to display comparative values]
Fig. 5. Created by VJ Um Amel, 2012. Research tools by R-Shief, Inc.
[Run your mouse over the graph to display comparative values]
Fig. 6. Created by VJ Um Amel, 2012. Research tools by R-Shief, Inc.
Figure 7 offers a cumulative display of the data in Figures 5 and 6. In this view, it is clear that Hamdeen Sabbahi is most discussed among English speakers, and Amr Moussa follows his lead. Looking at the history of Twitter talk, one could argue that Moussa lost ground to Hamdeen, as they seemed to have appealed to similar voters.
[Run your mouse over the graph to display comparative values]
Fig. 7. Created by VJ Um Amel, 2012. Research tools by R-Shief, Inc.
The timeline below is designed to help readers spot the highest points in the graphs (nodes). Like the cumulative bar graph above, Figure 8 offers a total number of mentions the candidates received throughout the month of May irrespective of language. According to this more comprehensive account, the following are the main highlights:
• Following the violence in Abbassiya late in early May, Twitter talk on Shafiq increased.
• The live ONTV television debate between Moussa and Abou Fotouh on 10 May generated Twitter talk on both candidates. However without further sentiment analysis, it is difficult to determine whether more talk about a candidate indicates support or ridicule. Gauging by the Twitter talk activity following the event, one could argue that the debates impacted both Moussa and Abou Fotouh negatively.
• At the close of the ballot boxes on Thursday 24 May, Shafiq’s blue dot soared while Morsy and Sabbahi both came from behind.
[Run your mouse over the graph to display comparative values]
Fig. 8. Created by VJ Um Amel, 2012. Research tools by R-Shief, Inc.
This interactive table below allows you to read through over 2,000 tweets posted that included the hastag #egypreselex while Egyptians went to the ballot boxes 23 and 24 May. The purpose is to allow you, the reader of this essay, to sift through a small sample of the tweets yourself.
[Run your mouse over the figure to display comparative values]
Fig. 9. This is an interactive table that allows you to read through tweets posted on #egypreselex
May 23-24 2012. Created by VJ Um Amel, 2012. Research tools by R-Shief, Inc.
In the interactive pie chart (Figure 10), the most frequently used words in the data chart appear. This chart is intended to offer a birds eye view into the semantic signals found within the corpus of data sampled for this study. Unsurprisingly, the most used word in the data set is “Allah,” the Arabic word for God, the second is the Arabic word for “today,” the third is the Arabic name for “Egypt,” the fourth is the Arabic word for “the people,” and the fifth most word in the data set is “before,” the sixth is a transliteration of the word, “retweet,” and the seventh is the Arabic word for “organization,” the eighth is the Arabic word for “day,” the ninth is the word for “revolution,” and the tenth most frequent word in the data sample is the Arabic name, “Mohamed.”
While Allah, retweet, and Egypt are obvious signifiers, the more mundane words like now, day, and organization are noticable. They may signify urgency, action, and mobilization expressed in Arabic on Twitter in 2 million tweets over 40 days.
[Run your mouse over the graph to display comparative values]
Fig. 10. Created by VJ Um Amel, 2012. Research tools by R-Shief, Inc.
Conclusion
• Despite negative sentiments and protests against military rule, Ahmed Shafiq has consistently been the subject of social media engagement in volume two and three times more than his opponents. However, Morsy’s high percentage of positive sentiment also indicates a strong base of supporters.
• The television debate on 10 May 2012 seemed to be a turning point for the liberal versus conservative camps. On the liberal camp we saw the sudden rise of Hamdeen Sabbahi from the Nasserite party. The conservatives stayed conservative and out of the limelight during controversial debate. Twitter talk on Morsy, in particular, became quiet after the debates.
• There is a fleeting aspect to this type of ambient journalism. It has a great impact on mobilizing and alerting around real-time events. However, the trends can change at a dizzying rate.
• Arabic language dominates Twitter posts on hashtags related to Egypt. In the case of hashtag #Egypt, the candidates trend differently when their names are posted in Arabic than when they are posted in English.
• The candidates that will advance to the next round, Ahmed Shafiq and Mohamed Morsy, both had very little Twitter discussion about them in English. The candidates that appeared most in the Arabic tweets are the two that won this first round.
• From this social media data, it was clear that Shafiq was “leading the Twitter talk polls,” so to speak, for weeks now. He consistently stayed ahead of his opponents throughout the race.
• Despite the minimal social media penetration among Egyptians, in this case study, Twitter did function as an echo-chamber to events on the ground—from the violent clashes in Abassiya in early May, to the television debates on 10 May, to the shoe throwing attack on Shafiq on 23 May, all the way until the closing of the voting polls.
_______________________
[1] Egyptian embassies and consulates worldwide administered elections a week prior to official voting in Egypt for Egyptian expatriats.
[2] Miller, Greg. (30 September 2011). Social Scientists Wade in Twitter Stream. Science: Vol. 333 no. 6051 pp. 1814-1815.
[3] Introduced by Frederic Jameson in his book, Postmodernism, or the Cultural Logic of Late-Capitalism, the idea of a cultural logic signifies a process of people collectively using effectively identical assumptions in interpreting each other`s actions. In this instance, culture exists in a spatial system of interests.
[4] According to UNICEF’s statistics by country, the total adult literacy rate (%), 2005-2010 in Egypt is 66%.
[5] According to Social Bakers, among Internet users in Egypt, the total number of FB users is reaching 10.7 million, which translates into a Facebook penetration rate of 13.26%.
[6] González-Ibáñez, R., Muresan, S., & Wacholder, N. (2011). Identifying Sarcasm in Twitter: A Closer Look. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies.