This article is part of a citizen-media data-analysis project, a collaboration between RuNet Echo  and the Maryland Institute for Technology in the Humanities . Explore the complete article series on the All the Presidents’ Tweets  page.
Our targeted Twitter data sample of over six million tweets contains tweets in many languages , but those are not always indicative of where people are tweeting from. While we were predominantly interested in what Russians and Ukrainians were saying about their presidents, we were also curious to see where tweets originated. As we collected all the tweets containing the last names of the presidents of Russia and Ukraine in Russian (Путин and Порошенко), Ukrainian (Путін and Порошенко), and English (Putin and Poroshenko), geolocating them could be useful for observing where people were talking about them more.
Geolocating tweets can be tricky, as there are several ways Twitter lets users indicate their location. For one, a user can specify a Place—manually select a city or neighborhood using a preexisting list of locations from a Twitter menu. Then there's the Exact Location—a set of coordinates usually provided via GPS or cellular triangulation. Manual Place selections are usually made from desktops or other fixed-location devices, so if a user ends up traveling, the last selected location would still be reflected in the tweets, making this location method less precise. The geo-coordinates of the Exact Location though usually use a mobile device's GPS capabilities and are thus more exact, but this feature is disabled by default in Twitter apps, so users must change their account settings for it to be enabled.
In their study  of a sample of tweets from the Twitter Decahose (10% of all tweets sent globally each day) from the fall of 2012, Leetaru et al. found that on an average day only 1.6 percent of tweets had Exact Location enabled. According to Semiocast , who looked at a broader sample of tweets in mid-2012, the geo-coordinates feature was used in roughly 0.77% of all public tweets.
The image above shows all Exact Location coordinates from Leetaru et al.'s sample. Because the percentage of geolocated tweets is so negligible, studies relying only on these tweets will almost surely have a skewed view of the Twitterverse, especially over short periods of time. However, the map shows that in terms of geolocated tweets, Twitter exhibits strong geographic diversity, and most countries on the map have at least some geolocated tweets. The use of the Exact Location feature is, of course, also tied to mobile Internet penetration and smartphone use in various countries. All of this means that as a single source of data, geolocated tweets are not representative, but as one of many facets providing information about a sample of the Twitter stream, they can provide some interesting insights.
The number of geolocated tweets in our Poroshenko/Putin dataset was 31013, or 0.49% of 6,342,294 total tweets we collected. This is somewhat smaller then the percentage of geolocated tweets in the larger samples discussed above, but still in line with the general trend.
All the geolocated tweets from our targeted sample are represented on the interactive map below, with a dot for each tweet. The saturation of particular areas with tweets is color-coded, so areas with more tweets are more red then yellow, and areas with a few tweets are more yellow than red. You can zoom in and out on the map to see the location of tweets in a particular geographic area in more detail by using the “+/-” controls at top left. Clicking on each dot reveals a pop-up window with an active link to the tweet in question.
From the visualization, it's obvious that North America, Western Europe, and parts of Eastern Europe have the largest share of the geolocated tweets in our dataset, and some parts of South America are covered as well. This is fairly reflective of the overall global geolocated tweets distribution from larger studies with non-targeted samples, with the exception of Asia, which is overall more active with geolocated tweets then in our sample.
The map  also allows you to zoom in on particular locations by searching for them in the search bar at top right. Here are some comparisons of geolocated tweet density of some key cities based on our sample (all magnified to the same level of detail).
The relative density of geolocated tweets in some of the cities above roughly corresponds to Leetaru et al.'s ranking of the top 20 cities  by percent of geotagged tweets in their 2012 sample. In their ranking, New York City comes second, Paris is fifth, London seventh and Moscow 20th. Given the theme of our targeted sample, it would be logical to expect more tweets about Putin and/or Poroshenko from Berlin, since Germany is closely involved in the Russian/Ukrainian politics at the moment, but Berlin users do not seem to be in the habit of geotagging their tweets.
Kyiv does not exhibit geolocated tweet density comparable to Moscow, but this is not surprising given the difference in population size  and the difference in smartphone penetration rates  in Ukraine and Russia. New York has relatively fewer geolocated tweets than Paris or London, perhaps mirroring both its geographical and its political distance from the Russian/Ukrainian political affairs.
While not entirely representative of the entire six million tweets, this smaller geolocated dataset presented as a heat map gives us some sense of how the world's attention to the two presidents and their politics is distributed.