Art from Twitter Streams?

Art generated by a computational device has been a topic that has been supported or decried for years. From its earliest beginnings in the 1960’s (Desmond Henry’s machine generated art) through to AI generated art impersonating human artist endeavours in the more recent past, art generated by computers has been the subject of much debate. Questions such as can computer programming code be considered art?, arguments between ‘traditional’ and ‘digital’ art, and how can we discuss the history of digital art as opposed to the well established history of art field?

Whether you are on the side of computer art being art or not, it is not the intention here to argue for one side or the other. It is though, the intention to present generated art from social media streams.

These streams are created when social media users self-classify and categorise messages using ‘hashtags’ for easy searching – bearing in mind that hundreds of millions of messages are posted each day on a social network such as Twitter. The hashtag makes it easy for a, for example, Twitter user to filter out all of the overwhelming noise to isolate a relevant stream of messages. These hashtags are not curated though, meaning they are user generated. There is no ‘standard’ hashtag to use for a particular stream. Some stream subjects suggest possible hashtags to use to curate discussion and debate – for example television programmes suggest a hashtag to use in messages to concentrate and collect debate into one stream. There is nothing to stop a user from adding their own hashtag to a message though.

These stream images have been generated using R. The images represent a simple sentiment analysis of the tweets (using the stock Syuzhet R package). There is no comparison to be made between the images as different colour palettes are used. They transform the hashtag streams to be viewed from a statistical, quantitative point of view to that of a cultural, even an artistic, point of view. The streams can be viewed using different colour palettes and different representational structures, which means there is no one way to represent a stream.

Given all that though, we can look at how images of a stream changes over time when using the same colour palette. For example, the #ge2019 stream:

#ge2019 taken at 0930, 12 December 2019 as the election was happening

#ge2019 taken at 2230, 12 December 2019, after the polls had closed

Viewing the streams artistically, I would say, is very different from viewing the streams from a quantitative viewpoint. Any comparisons are subjective and potentially masked by the representation and structures used to display the stream. However, with the huge volume of data generated by social media streams, then their artistic representation holds much potential.

TVX 2014


TVX2014, or as a longer winded version, “ACM International Conference on Interactive Experiences for Television and Online Video 2014” was recently held in Newcastle, UK. Newcastle is the party capital of the UK (cue many arguments…), so it was a lively conference in a lively city!

Our long paper, “Disinhibited Abuse of Othered Communities by Second Screening Audiences” (links: eprints, ACM) was a look at how TV viewers treat participants and stars on TV shows. In particular it looked at how the TV show “Thelma’s Gypsy Girls” was commented upon by a second screening audience. We investigated the tweets that were posted during the show and that contained the “#thelmasgypsygirls” hashtag. When TV shows encourage a discussion of the content and issues that are broadcast through the showing of a hashtag on screen they have no control over that discussion. If it took place in the broadcast, or if the shows producers had some kind of ownership of that discussion then certain ethical and other guidelines must be adhered too. As it is, as the discussion takes place on a completely separate platform and the producers and broadcasters have no control or ownership of the that platform. The discussion falls through a loophole in the ethical guidelines governing broadcast TV and there is no control over it (apart from the platform guidelines, such as Twitter’s terms and conditions).

It seems strange that a TV show is encouraging discussion and debate over its content, yet it has no control over that debate. Freedom of speech and liberal censorship and libel laws allow these discussions to take place with only legal repercussions if the debate gets out of hand.

TV and Twitter are uneasy bed fellows. On one hand, Twitter sees TV and the discussions which take place around TV shows as vital to its growth and usage in the future. TV shows  see Twitter as an important place to host discussions about their issues or content. However, with no control over that debate, it seems that TV shows and the broadcasters are selectively myopic when it comes to these debates. Sure, they can give instant feedback and comments about the show, however, they can also be hate fuelled and libellous towards individuals or groups within the shows.

The 1971 classic, “Le Mans” is only very occasionally shown on UK TV. Starring Steve McQueen, the film is considered a benchmark in racing films. McQueen raced this car with Peter Revson (the 48 Porsche 908/2 Spyder) in the 1970 12 Hours of Sebring race, where it came first in its class, and second overall.

Inside the F1 twitterverse. Part 1: The Teams

[flickr]7774915380[/flickr] Have you ever wondered what the world of Formula 1 twitter looks like ..? No? Well maybe you should. The use of twitter as a social communication space has embraced the widening of its use to, amongst other things, augment TV broadcasts, live event consumption and now it allows fans of any sport you could mention to follow their favorite teams and stars with a 24/7 feed of information and updates. Formula 1 has been at the forefront of social media channel exploitation to give fans access to the rarefied world which exists behinds the scenes in this global sport. All twelve teams have adopted twitter as part of their social media communication strategies – with varying degrees of information and levels of communication being employed. Generally, the team accounts provide a top level communication in a simple hierarchy, with driver and team member accounts completing the landscape with individualized information for fans and interested followers.  In this post we look at the teams and how followers are distributed among them. An analysis of how drivers fare on twitter will follow.

Teams and followers

Each of the twelve teams have a twitter follower-base which range from a modest 47,000 for Toro Rosso, to  316,173 for Ferrari.

The number of unique followers in these totals is 670,300, which doesn’t sound like a lot when it comes to a global sport such as F1. When analysing twitter user numbers when related to large scale phenomena such as sports fans and TV show audiences it becomes apparent that twitter, while generating vast quantities of information and data, actually only accounts for, in the case of TV audiences a small percentage (< 5%) of the viewing audience. This effect is more pronounced for F1. With a global audience figure of around 520 million (, the team’s twitter followers reflect only a tiny 0.12% of the global audience.

What is the behaviour of the followers of F1 teams?  If we look at the number of teams followed by each twitter user, the results show an interesting distribution:

With 68% of the teams followers only following one team, does this underline a partisan quality to F1 supporter behaviour? Certainly supporters on the tribunes across the world show their support for their chosen team in this way, be it the Tifosi of Monza or the rocket red caps at Silverstone. A closer look at this behaviour using the social network analysis tool, ‘Gephi’, reveals the dramatic nature of this partisan follower behaviour:

The clusters are formed from the single team followers, with the links between the teams being formed from the followers who follow more than one team. Each connection or ‘edge’ represents a ‘follow’ by a single user to a single team.This network only represents a 1% snapshot of the F1 teams twitter network – computation times and data manipulation issues on the MacBook became too significant when handling larger data sets. It does, though, give a graphic indication of the nature of F1 teams follower behaviour.

If we look a little closer at the followers who follow only one team – arguably indicating a defined level of support or partisan following, the results give an indication of the type of support each team receives:

With a level of 56% of their twitter followers following only them, Ferrari, unsurprisingly, has the highest level of partisan support from twitter followers. McLaren and RBR follow with significant levels of core support from twitter followers. A surprising bit of information from this data is the status of the Caterham F1 team. While enjoying a top 6 level of support from twitter followers, the proportion of those following only Caterham is low at 14.7%. This possibly reflects its status as a relatively ‘new’ team, but having a majority of followers who also follow other teams. Are Caterham everyones favourite ‘second team’?

The data which twitter produces is vast and almost endless in its research potential. Certainly the perspective it allows here is interesting. The defined segmentation of the supporter base by the identification of single team followers is interesting for teams, their sponsors and F1 in general. The segmentation is probably not anywhere near as  defined as, say, that of Premership football teams, but it still offers an insight into the reach and exposure to core fans that twitter offers.

What a twitterific jubilee..!

Yesterday’s Diamond Jubilee River Pageant was a magnificent occasion. As is usual now, the world of twitter was alive with TV viewers discussing, ranting, or just observing the proceedings as they watched. The use of twitter and other social media channels while watching TV is rising in popularity, and it is generating vast amounts of data for researchers. Viewer opinions, thoughts and instant feedback are available to be searched and mined for new perspectives on how we consume our TV entertainment and how we interact with friends and others while we are watching.

The #bbcjubilee hashtag was a focal point for viewers of the BBC coverage of the Pageant, and I collected over 46,000 tweets during the broadcast. This is a reasonable amount, but doesn’t come near the  volumes of tweets produced during XFactor broadcasts for instance – probably reflecting the different viewing demographic and the type of programme.

The wordle above shows the relative occurrences of the 150 most popular words in the #bbcjubilee tweets. As you would expect, the Queen and members of the Royal Family feature prominently, but then as we have seen in other TV related streams, the BBC presenters and other celebrities feature also. The phenomenon of celebrities tweeting about the broadcast and being retweeted in large volumes is a common observation and is seen here with comedian, Frankie Boyle and Radio 1’s Matt Edmondson featuring.

The network of viewers discussing  the proceedings shows a typical profile for TV audience Twitter networks. Large numbers of users linked to one or two others through ‘mentions’, while a small number of celebrities and popular tweeters are mentioned by large numbers of other users as their tweets are retweeted through the network, seen here by the large clusters in the centre of the network.

 This gives us more good data to try to understand the social nature of these networks of, in most cases, non co-located TV viewers, and what it can tell us about TV audiences in general. I’m looking forward to discussing this with others at the upcoming EuroITV 2012 conference in Berlin!

Academic year ends – next years challenges?

[flickr size=”large”]6998431820[/flickr]

We have had a busy year in the School of Computer Science this year. As well as securing BCS (British Computer Society) accreditation for our undergraduate and postgraduate taught courses, we have hosted a Microsoft Windows phone 7 ‘phone camp’, a 24 hour ‘Game Jam‘, and held numerous open days and applicant taster events.

The challenges for next year will  not only be to maintain these achievements and to further improve the student experience in the School of Computer Science, but also to adapt to the new fee structure within which our new students will be studying. Contrary to the popular media-supported belief, these students will not have to pay the course fees up front, but will start to pay them once they have graduated and are earning above a threshold of £21,000. Will this new intake of students coming to University under this new fee regime have different expectations and demands to our previous intakes? That is difficult to say. What is clear however, is that the course standards and requirements will not change – there may eventually be more transparency in how courses are administered and managed however. This is already the case with External Examiner reports for example, as they are considered public documents, as are QAA institutional review reports and other documents. The evidence is already showing that the application pattern has changed slightly this year, with students deciding on their final University choice later in the application cycle.

Research challenges are growing both in the School and across the University sector as the 2014 REF deadlines get ever closer. Research outputs and publications are high on everyones list of priorities. I will be in Berlin in July at EuroITV 2012 to present my latest paper “Who is on your Sofa?  TV Audience Communities and Second Screening Social Networks“. I presented a short paper at this conference last year in Lisbon and found the conference to be a fascinating mix of computer science, HCI, TV broadcasting and technology. I expect this years conference to be just as good as last year!