Is Snowden's international popularity a myth or a reality? Global Social Media Monitoring Results

Edward Snowden does not exist and is spoken of only in Russia. “Snowden is a traitor, an enemy and a traitor,” say US residents. “Snowden is an international hero and an example for imitation,” they say ... well, somewhere they probably say, we thought. When we started to work on a new task, it seemed to us that the topic of Edward Snowden was intrigued by the whole world. Therefore, the task set by our partners from the Vox Populi Foundation, which specializes in the study of public opinion in social media, seemed to us rather simple: to assess the interest of the population in the situation of Snowden in the world, and in Russia and the USA in particular. But social media monitoring for sociological research is a new entity, and therefore doubly interesting: first, you never know what kind of result you will get; secondly, you admire the possibilities of the social universe created by mankind. The result and this time turned out to be somewhat larger and rather unexpected: we analyzed a multilingual stream of messages from 230 (!) Countries of the world. About how we divided by language and geolocated this kingdom of Babylon - under the cut.

Geolocation messages

Over the past year, we have repeatedly conducted studies on different countries of the world and for different languages. But it was always all the same that was one or several neighboring countries: French opinions were selected for analysis in France, Spanish-language messages were selected for elections in Venezuela, etc.

Even then, the difference in the mentality of people in the use of social media became obvious: for example, for the French, the percentage of geolocation (geolocation) of Twitter messages is 75%, for Venezuelans 80%, and in Russia only 44%.

An initial analysis of the situation with Snowden showed that interest in the topic is global and multilingual. So, we could, in practice, assess the merits (or disadvantages) of the new, expanded geolocation and language autodetection modules, the development of which was used to devote considerable forces to developers and linguists.
Why is it important for us to understand the language of the message during geolocation? The thing is that to determine the location of the author, we analyze not only the profile data, but also his posts, tweets, comments, etc. That is, we take into account the geo-data from the author's profile, update them with each new message. We take into account the geo-tags to the messages of the author. We consider geo-information inside messages. We make a bunch of the author's profile with his profiles in other social networks, and take into account the geo-data there. We run it all through our geo-dictionaries in order to bring data from different formats indicating geo to a single species and make a bundle of city-region-country.
')
The theme of Snowden allowed to explore the "temperature in the hospital" for the geolocation of social media for the whole world. For the analysis, Twitter and Facebook messages were used, plus VKontakte for Russian-language messages. Result: geolocation is determined for 57.8% of all authors at the country level and 41.4% with accuracy to the city.

Factualization of the study (data from 1 to 20 August):

- over 600,000 messages;
- 230 000 unique authors;
- from 230 countries of the world (8 000 settlements)!


This is such a surprise, I think, for many of us, there are more than 200 countries (and territories) in the world. For example, there are such beautiful names: Wallis and Futuna, Saint Lucia, Bouvet, Mayotte, Kiribate and even Niue :-)

To analyze the messages, spelling of the Snowden name in different countries was used:
USA, England, Germany, France, Poland, Czech Republic, Hungary, Romania, Moldova, etc .: Snowden
Bulgaria: EDWARD SNOWN
Belarus, Ukraine: Edward Snowden
Kazakhstan: Edward Snouvden
Slovakia: Edward Snowden, Edwardovi Snowdenovi
Macedonia: Edward Snowden
Serbia and ex-Yugoslavia: Edward Snowden, Edward Snowov Albania Edvard Snovden
Azerbaijan: Edvard Snouden
Greece: Σνόουντεν

The topic turned out to be very interesting for our linguists. In addition to checking the language definition module (17 languages) and geolocation, it was necessary to find a practical specific solution for toning not just short texts (this task was solved several months ago), but also very, very slang expressions that manifest themselves primarily in tweets, for example :

U r wrong Obama. Snowden is a patriot. It is u who r the traitor to the united states
Source: twitter.com | by Jim Hackney | USA, Springfield | 14.08
21:05:34

The pool also included messages in Arabic, hieroglyphic or local languages, for example:

لاق خدمة avلار دمة
Source: twitter.com | by TAlbahussain | Double: 49 Saudi Arabia, Riyadh | 10.08 18:46:52

Quyêt định chờ xem mưa sao băng. Haiz, lão Snowden rât "ang yêu" nha, muôn giêt qua B-)
Source: facebook.com | by Kim Eun Won | Republic of Korea, Seoul |
12.08 18:45:38

RT @supinya: กรณี #Snowden ปล่อย ข้อมูล ว่า สหรัฐ สอดส่อง โดน โดน ไล่ ล่า แทบ แย่ เพราะ ไม่มี รัฐ ใด
ยอมรับ ว่า สอดส่อง จริง แต่ ด้าน ตำรวจ ไทย ภูมิใจ นำ เสนอ เสนอ มาก แปลก แต่ จริง
Source: twitter.com | by Natdanai | Thailand, Phuket | 13.08 15:36:58

Such messages were used to assess interest in the topic, but were not toned (not conducted through the message tonality determination module).

The results of the study on Snowden published in the "Kommersant" .

It is a pity that we did not have time to expand our map module on the world map, which was sharpened for the regions of Russia, the USA and Europe.

We will work on cartography and on a whole range of new tasks that our first “world” social media research has revealed. How will something new-interesting - be sure to publish in our corporate blog. See you again!