Visualising Twitter Networks: John Terry Captaincy Controversy

Visualising Twitter Networks: John Terry Captaincy Controversy


After our recent CAST Lab session ‘Web hacks and visualisations’ with Tony Hirst, on the use of Google, Twitter and Yahoo Pipes APIs and Gephi to scrape and visualise social networks, I decided to use these methods to explore how many people were tweeting about John Terry in the newly ignited row about his prosecution, loss of England Captaincy and the resignation of England manager Fabio Capello in February 2012. I used a combination of Google Docs Spreadsheets, a formula to pull in a Twitter API search for tweets about John Terry controversy, and later exported this into data visualisations tools like Gephi to try and visualise a network of the most followed and tweeted at people on Twitter.

Now, I wanted to come at this task purely from the perspective of a Social Scientist, who is using Digital Social Research Methods to investigate the social network from within Twitter, and not, per se, investigate what I already knew what was going on in the media with regards to the John Terry racial abuse controversy.  With this in mind, I deliberately ignored all media coverage of Twitter conversations around the key actors in the controversy for the 4 day period in which I did my research, and instead decided to use the research tools and methods I had learnt in the CAST London lab to explore the Twitter network and this controversy from the inside-out.  I wanted to see if I could indeed find a pattern that I could follow and unravel with further investigation, with my end results hopefully leading me down a gingerbread trail into what was really happening and being reported by the media in more traditional circles.


Following instructions from Tony Hirst on how to import dynamic data from Twitter using both Twitter and Google’s Dynamic API, I used Google Docs to publish data of the top conversationalists on Twitter who mentioned John Terry over a four-day period between 3-7th February.

I first registered as a developer on Twitter, and then confirmed my Twitter Authentication code and consumer keys:


Then I went into Google Docs and using the ‘=ImportData(‘ formula , ran a search for all tweets related to John Terry, which gave me the results below.  I have embedded sections of my Google Docs feed into this post, so that you can explore the data.

The amount of data I retrieved was vast, and included actual tweets, account descriptions and id images (I knew peripherally that the data was so large due to the immense media attention on the whole John Terry vs Anton Ferdinand controversy, and on later investigation, the added element of England Coach Fabio Capello’s commentary and then resignation of his post over the FA’s decision to strip John Terry of his England Captaincy ahead of the outcome of any Civil proceedings later in 2012.):

I wanted to filter down this data from what would generate over 4000 nodes and connections, to just the top conversationalists.

To filter this, I extracted the data from the ‘from_user’ and ‘to_user’ columns into another Google Docs spreadsheet, with the aim of exporting this as a CSV file and loading it into an API to visualise my network. This selection of data did not include re-tweets with ‘@*’, as this would be a misrepresentation of actual conversations, where users are only re-tweeting whole messages and not really taking part in a two-way communication:

While reading Tony Hirst’s Ouseful Blog, I came across his mention of Martin Hawksey’s TAGSExplorer tool.  Hawksey has written an interesting post about how to ‘archive event hashtags and create an interactive visualization of the conversation‘ through Google Docs in a slightly different way to Tony Hirst, with the use of this Google Spreadsheet template. I tried out Hawkesy’s method out of curiosity – Hawksey was really helpful with a few glitches I had along the way, and I managed to extract a good dataset.  However, I found that it gave me very similar results, so decided to stick with the data that I had collected previously. What I also found really really useful about Hawksey’s post, however, is his TAGSExplorer tool, which allows you to paste in the link to your Google Docs data and quickly visualise the connections between the nodes and networks that you have collected. I decided to give this a try and came up with some really interesting results.

Because this tool is very popular, and can crash from time to time, I decided to make some screengrabs of my findings, as well as put a link to my network (it may be possible to see my network in full animation, but just in case the system is down, please see my screengrabs and slideshows):
Link to my TAGExplorer network:

TAGSExplorer Screengrabs: The main network of conversations mentioning ‘John Terry’:

TAGSExplorer has a very useful set of filters, one being the ‘Top Conversationalists’ tab, which showed me that the top 4 conversationalists were: joey7barton, stancollymore, rioferdy5, and frankieboyle. Now bearing in mind that I have hardly any knowledge of footballers, football commentators, or teams, these names seemed marginally familiar (perhaps I had heard mention of variations of them in the press somewhere), but unless I actually Googled them, I had no idea who these top conversationalists were. However, just by looking at the list below, by far the biggest network was for an account owned by user ‘joey7barton’:

It wasn’t until I had Googled the username and found that this account belonged to Queens Park Ranger Captain and Footballer Joey Barton, that ‘all the dots (or ‘nodes’) started to add up’ so-to-speak).

Further Google research found that the reason why Joey Barton’s account (joey7barton) was tagged by Twitter as a ‘top conversationalist’ in relation to tweets about John Terry wasn’t just because he was a famous footballer with a large following of over 1 million users, it was because in the days leading up to the re-ignited John Terry vs. Anton Ferdinand racial abuse row, Joey Barton himself had been tweeting his robust views on the whole case, from his perspective of being on the pitch when the incident happened in 2011 – he had also tweeted in response to being reprimanded for tweeting about the incident and being in possible contempt of court, due to the possibility of him being called as a material witness in the coming trial (MadNews, 5th February 2012).

Suddenly, this use of Google Docs, the Twitter and TAGSExplorer API turned from an interesting exercise into Social Network analysis using Digital Research Methods, into an eye-opening journey into discovering how Twitter’s Top Conversationalists discussions about John Terry connected them as key sporting celebrities and commentators in the public eye. These celebrities were all commenting and contributing in their own way to this high-profile case, illustrating a complete a change from the older traditional ways of celebrities expressing their views by writing a letter to a newspaper or appearing in a TV or radio interview – through the use of digital social networks, they were actually engaging in live conversations with members of the public on this topical issue of the day.

Further investigation found media articles mentioning some of my top four twitter conversationalists:

Not surprisingly Twitter is buzzing with the news that England manager Fabio Capello has resigned. Both current and former players as well as journalists have expressed their opinion on the Italian’s shock decision.

Here’s a selection of the tweets that have been sent so far, after the announcement.

The question everyone is asking, fans and players, where do we go from here? Euro’s is just around the corner and we have no manager?!?

Stan Collymore
England fans and players need a lift.A motivator.A man with experience. We’re not going to out play or out coach Spain in 2012. #Harry

Just spoke to 2 current England Internationals on the phone. They are shocked but not surprised,and not unhappy at all.

No player was slagging Capello.Just merely acknowledging a few issues were not resolved on and off pitch over 2 years.

Joey Barton
Well, well, well, this debacle claims yet another victim. Where does this stop. #madness

No captain and no manager. 4 months from a major championship. What’s going on…….

Rio Ferdinand
So Capello resigns….what now….

Source: MSN UK Sport: Fabio Capello’s resignation sets Twitter alight (8th Feb 2012)


It seemed that my simplified network was an effective way of visualising the top-level conversation between footballers and commentators.  With this new perspective, I went back to TAGSExplorer to investigate this further.  I used Quicktime to create a screen-recording of archived playback of twitter conversations between members of the public and top conversationalist Joey Barton. These reflected the reports in the media, but by playing back these conversations, gave these reports a more visual and unfolding view of the network as it formed. I have placed screengrabs of that recording in the Flickr Slideshow API below:


In my investigation, I also happened to click on another user mysteriously called ‘Frimpong26AFC‘.  The playback of conversations with this user’s account seemed interesting, with a mixture of defence of John Terry as not being guilty until tried, to who should replace him as England Captain.  Again, having no real knowledge of English football apart form the main contenders in this saga, it was interesting to see from a Google search, that again, this top conversationalist was another FA Premiership footballer, Arsenal Football Club Mid-fielder, Emmanuel Frimpong.  See the Flickr API below, which shows his Twitter account, and the conversation I happened to listen in on:


However, any further investigation was unfortunately halted due to increased difficulty with time-out failures on both Google Docs, and TAGSExplorer.  The media and Twitter storm over Fabio Capello’s comments and then resignation over the John Terry controversy, meant that the APIs sourcing both applications slowed right down as they tried to pull data from Twitter.

But I wanted to delve into this further, even if it was just a deeper investigation of the Twitter network I had archived.  So, realising that  I could use the skills I had learnt in the CAST Labs to visualise my own story of the Twitters Top Conversationalists in the John Terry saga – I resorted to importing the data I had archived from Twitter into Gephi and exploring how I could visualise the links within this network from there.

Having imported my data into Gephi, I also imported a few more plugins like Parallel Force Atlas to improve the layout and positioning of the nodes, and used network statistical analysis plugins like the Eigenvector Centrality and Connected Components plugins to calculate and visualise key sub-networks that would more clearly show the Top Twitter Conversationalists.  I used the Nodes Ranking and Closeness Centrality colour filters to highlight the key conversationalists – a screenshot of this is shown below:


However, I wanted to show this in a more interactive way, so looked at the tools I had in Photoshop, to see if I could animate this somehow. In the Photoshop menus, I came across the ‘Export to Zoomify’ feature, which enables you to output your large image to a zoomable image on the web. I couldn’t get this to work straight away, so decided to find out if the ‘Zoomify’ feature had been upgraded in any way. Thankfully it had been developed into a web service called ‘‘, which allows you to upload a high-resolution image that the app will turn into a zoomable object. The result of my uploaded screengrab of my Gephi network is below. You can interact with the image by hovering your mouse over it, and zoom in and out, and click and drag to pan around the image to get a closer look at the top conversationalists, for example, footballer Joey Barton (Joey7Barton):

However, what troubled me about the network above, is that apart from looking at the larger clusters, it was not so easy to see the difference in the top conversationalists, and those with lesser tweets. To better see what was going on, I did some screengrabs of the top Twitter Conversationalists, and imported these into a Flickr Slideshow app, which I have embedded below. Again, you can see the top four conversationalists are: Joey Barton (@joey7barton, Stan Collymore (@stancollymore), Rio Ferdinand (@rioferdy5), and Scottish presenter and commentator Frankie Boyle (@frankieboyle) – with some interplay between his account and radio sports show Talk Sport Drive (@talksportdrive). I have provided screenshots of the actual Twitter accounts so that you can see these are coming from the offical accounts of these footballers/commentators:


However, it was still hard to see at a glance of the whole network, the clear differences between Twitter Top Conversationalists and those with smaller connections. To correct this, I went back into Gephi, and used the ‘Betweeness Centrality” algorithm to put emphasis on the links between people tweeting about John Terry.

Doing this, and adjusting the colour-scheme slightly makes it much easier to spot who the top conversationalists are. They are coloured in pink, and have been moved to the left. If you zoom-in, you can again spot Joey7Barton’s network in the centre of the left-hand part of the network, and Scottish commentator Frankie Boyle, to the bottom-left central left. Footballer Stan Collymore has a strong network in the top-left, and sports radio talk show ‘talkSportDrive’ is just below him. In the bottom left, you can see that ‘rioferdy5’, footballer Rio Ferdinand (Anton Ferdinand’s brother) is also a top conversationalist – this mirrors newspaper reports about how vocal Rio has been about the whole incident between John Terry and his brother Anton (Telegraph, 2012). The paler colours show conversations where only one or two tweets have occurred, and the user does not have a large amount of followers on Twitter.  You can explore these newly defined nodes more in the of my network below:

Results and Discussion:

What I found really interesting from my investigation into the John Terry controversy, was that when I came at it from the point of view of a pure novice, who doesn’t know a thing about football, and had deliberately not read any media articles about the people tweeting about the saga – was that I came across a whole new network of ‘Top Conversationalists’ who were actually made up of professional and ex-footballers, all engaging in discussions with the public about the controversy, and all creating a clearly visible network of this once I mined the Twitter data with a combination of the Twitter API, Google API, TAGSExplorer, and Gephi. Using these tools as my guide, and then on the final steps using Google Search to find out just who these ‘Top Conversationalists’ were lead me to discover that the world of Twitter, football celebrity and media commentary can be made visible and linked in a way that dos not require prior knowledge of football or commentary. What is also interesting is that these tools are showing a level of communication between celebrity and public that was not so present before the age of digital social networks.

In his discussion of ‘bottom up leadership-follower patterns on Twitter’, Geser (2011) argues that Twitter gives rise highly asymmetric leader-follower patterns (Geser cites Niles 2009). Thus, much influence concentrates on the two percent “leaders” able to accumulate 1000 followers or more (Geser cites Sysomos 2010). This was shown by the large amount of followers the top conversationalists in my network had, and how much this seemed to influence discussions around controversial topics of the day.

Geser illustrates this point further by arguing that effective Twitter leadership may well be based on offline factors: e. g. on the incumbency of a high political office or on a charismatic celebrity reputation acquired in sports or entertainment – a factor shown in my discovery of sportsmen and commentators as top conversationalists. Twitter also seems to act as an agent of freedom, where celebrities can speak thier mind without the constraint of thier managers or agents, and ordinary citizens have direct access to celebrities (Geser 2011) – although in the case of Joey Barton, can also get them into trouble, where his tweets on the John Terry controversy led to a possible contempt of court (MadNews, 5th February 2012).

For future research, I would like to use a combination of Python, R and Gephi to investigate who the real contributors really are behind the tweets of these top conversationalists, and have done some preliminary investigations based on Tony Hirst’s ‘Mapping Corporate Twitter Account Networks Using Twitter Contributions/Contributees API Calls’ post, which led to some discussion of the tools and libraries I could use (see comments on Hirst’s blog).

I would  also like to log the use of hashtag activity over a longer period of time, and upon further discussion with Tony Hirst, install ThinkUpApp on a server to see if I can dig down further to get more in-depth results, and explpore Martin Hawksey’s use of an online gexf viewer for visualising gexf files generated from Gephi.


ESPN Soccer. ‘Fabio Capello quits as England coach’. Acessed 09.02.2011:

Geser, H. (July 2011 (Release 1.0)). “Has Tweeting become Inevitable? Twitter’s strategic role in the World of Digital Communication.” Sociology In Switzerland: Towards Cybersociety and Vireal Social Relations. Accessed on 09.02.2011 from:

Hawksey, Tony: How to archive event hashtags and create an interactive visualization of the conversation. Accessed 04.02.2011:

Hirst, Tony: Ouseful Blog: Mapping Corporate Twitter Account Networks Using Twitter Contributions/Contributees API Calls. Accessed 02.02.2011:

Hirst, Tony: Ouseful Blog: How to import dynamic data from Twitter using both Twitter and Google’s Dynamic API. Accessed 02.02.2012:, 5th February 2012: UK NEWS: FOOTBALL STAR JOEY BARTON INVESTIGATED OVER JOHN TERRY TWEETS. Accessed 05.02.2012:

MSN UK Sport. Fabio Capello’s resignation sets Twitter alight Accessed 08.02.2012:

Telegraph (2012): Rio Ferdinand’s animosty towards former England manager Fabio Capello is not ‘lost in translation’. Accessed 10.02.2012:




  1. Tony Hirst

    Hi Sam – wonderful write up; it’s great to see how you engaged with the data and worked with it to develop a view over the controversy. I feel your pain in finding effective ways to display large graphs in meaningful way online. Marthin H has explored the used of an online gexf viewer for visualising gexf files generated from Gephi ( ) but this is still very much an active area of development I think.

    If you want to log hashtag or search terms activity over an extended period, I recommend you look at Martin’s TAGSExplorer again, or if you have access to a server, maybe consider something like ThinkUpApp.

    I’m looking forward to see what you explore next…;-)

  2. Sam

    Hi Tony,
    Thank-you for your kind words, and for your input – Martin Hawksey’s exploration sounds very interesting indeed, and installing ThinkUpApp on my test server sounds like a good way to search terms over the extended period I am considering. I’m really enjoying using this digital research methods, and definitely look forward to exploring more in the future. 🙂

  3. Pingback: OER Visualisation Project: The heart and pulse of #ukoer [day 40] – MASHe

  4. Martin Hawksey

    Hi Sam – Your experiences with TAGSExplorer made me think if the backend is processing the spreadsheet to essentially make an edge list why not have an option to output this for using offline in tools like Gephi. So now if you add &output=true to a TAGSExplorer url you get a comma separated edge list to use with you SNA package of choice (I should really call it Sam’s list ;). I found it easier to import the edge list in Gephi’s Data Laboratory tab. I need to document the process. You can see my attempt at processing a conversation graph here


    1. Sam

      Hi Martin,

      Really great work-around, thanks for that – I am now effortlessly outputting data from my TAGSExplorer sheet. 🙂 You’re right, Gephi’s Data Laboratory tab is much easier to use – I also find the ability to pin-point specific nodes to then view in overview very handy. Thanks for the link to your article – some interesting reading, and I like your use of Google presentation – hadn’t realised it had come along so much – I look forward to trying it out soon.


  5. Pingback: Data Visualisation « AgNetPro

  6. Pingback: Mapping Twitter Abuse and Recursive Abstraction: Can We Accurately Map Who Abuses Who? « Twitter Abused

Comments are closed.