presentation: big data + twitter

By On 2013/04/27 · Leave a Comment

Earlier this week I gave a presentation at the sixth Perth Georabble, as part of the international Big Data Week. The Big Data theme provided some scope to cover topics less directly relevant to locative or spatial data and analysis, and I was invited to give a short talk on the various Twitter mapping projects we’ve been carrying out at QUT and Curtin. Below are my slides, although no audio this time around – hopefully it made for a clear outline of our work and the various datasets and types of analysis we work with! There were some really interesting presentations given that evening, covering questions and topics around Big Data around different institutions and projects, including a fascinating overview of the Square Kilometre Array and its data (and the quandries they raise), and reflections on data collection from shipwreck sites for the WA Museum. My thanks to the organisers for inviting me, and hopefully I’ll make it back to the next Georabble in a month or so.

 

 

(The slides are also available through the Conference Slides section of this site; you can find more information about the various projects mentioned in the slides by following up the publication details on the relevant slides, or in some cases having a look through my other conference slides and publications)

 

of mice and memes

By On 2013/04/06 · Leave a Comment

Some of the research I’m preparing for more thorough analysis later in the year builds on what I have been studying, along with my CCI and Curtin colleagues, over the last few years regarding social media uses and behaviours. Many of our presentations and papers have covered isolated events or datasets, across a range of contexts, and these have provided very exciting and important new information about how Twitter in particular is employed for live commentary, for political discussions, and for crisis communication, to name a few examples. One of the research themes I’m keen to follow up this year, though, is taking a step back from these isolated cases, and exploring the wider patterns and behaviours that can be identified across multiple datasets. Some of these ideas I’ve tried teasing out in conference abstracts, although ultimately these may not be successful submissions in terms of being accepted for presentation!

One of these themes is the development of conventions mixing social media commentary, breaking news, and satire, through the creation and cultivation of parody and spoof accounts on Twitter. Here, I am not referring necessarily to accounts which are impersonating other people with a deceptive air – deliberate imposters – but to accounts which are acknowledged fakes or have clear humorous intentions. There are many different examples of these accounts, which satirise everything from public and historical figures (the @Queen_UK account, for example) to fictional characters (@Lord_Voldemort7) and organisations (@DeathStarPR, representing the public relations wing of the Galactic Empire from Star Wars). Further accounts appear as the tweeting voices of towns, animals, body parts, or artworks. Some of these accounts are long-running, established parts of the Twittersphere, and tweet – in character – about many major news stories and cultural events; in the research that Stephen Harrington, Axel Bruns, and I carried out around last year’s Eurovision Song Contest, @Queen_UK was a popular source of comments to retweet during the UK live broadcast (tweeting around the presented context of the monarch having a Eurovision party herself). Similarly, @Queen_UK and @Lord_Voldemort7 also appeared as prominent users within Farida Vis’s dataset of tweets around the 2011 London riots.

Other accounts, though, are more short-lived, and often context-specific. These accounts are often created in response to breaking news or, in particular, live televised events when unusual or amusing (and unplanned) developments transpire. In 2012, for example, a cat appearing on the pitch at a Liverpool soccer game immediately inspired an Anfield cat Twitter account, while a chicken invading the pitch at a Blackburn soccer game (as a fan’s protest) led to a number of ‘Ewood chicken’ accounts. When Clint Eastwood addressed an empty chair at the 2012 Republican National Convention, several Twitter accounts appeared soon after, claiming to be tweeting as the chair itself. The world of academia is also a supportive environment for such accounts; during last year’s Association of Internet Researchers conference, somewhat inevitably, a Twitter account appeared on the final day of the conference on behalf of a number of toy animals found in one of the presentation rooms.

Building on the studies that Limor Shifman has carried out into memes and viral videos on YouTube, I’m looking into these satirical, parody, and spoof accounts as a memetic practice – essentially, as an established trope of social media commentary. It is an expected behaviour that (new) parody accounts will appear within ongoing coverage of a variety of topics on Twitter, and when one appears, others may follow; in a forthcoming book chapter about tweeting during the 2012 Tour de France coverage, for example, I discuss in part the presence of parody accounts for many of the SBS commentators, and the creation mid-race of further accounts following developments within the Tour itself.

Of course, these satirical accounts do not necessarily aim for a wide audience or for longevity; their inspiration and focus may be local rather than global. For topical accounts, their activity may decrease dramatically – or stop completely – once the initial wave of interest has subsided (the many Ewood chickens mentioned previously mostly stopped tweeting within a day or two after the match in question). However, what is clear is that the creation of these accounts is now an established response to live events and breaking news, regardless of their scope or lifespan.

I want to stress the local aspect in this post, rather than the global reach of the likes of @Queen_UK and @DeathStarPR, since Perth in particular seems to present a environment supportive of parody and mockery on Twitter (other towns of course will also inspire this activity, but the Perth context is the one with which I am most familiar). Four years ago(!), I wrote about the spread of links to a video satirising life in Perth – entitled “This is Perth” – which attracted a rush of attention across Facebook and Twitter within hours of being uploaded. This satirical presentation of Perth can also be seen in many different Twitter accounts which have appeared over the last few years; Exciting_Perth, for example, or the accounts representing public artworks in the CBD, The Perth Cactus and The Perth Pineapple/Totem. As with other regional and national contexts, politicians are not immune to parody, and (as I noted in the WA election posts) the state Premier, Colin Barnett, is the subject of several spoof accounts (such as YourMateColin or Emperor Barnett). Wildlife are also represented; in particular, the prevalence of shark sightings during the summer (and autumn, and spring) has inspired the Perth Sharks account. Finally, of course, and pre-dating Twitter, the Worst of Perth website should be noted as a major source of mocking many different aspects of Perth life and culture (also on Twitter). Like Eurovision, in some ways, there is a degree of ironic detachment and anti-fandom when it comes to discussing Perth online.

It is somewhat inevitable, then, that further parody accounts will arise around Perth-specific themes. Earlier this week, for instance, photos, videos, and comments about rodent problems at the Hoyts Carousel cinema were posted online by cinemagoers; during the day, this story spread such that not only did mainstream media outlets cover it (and Hoyts was later ordered to investigate the problem further), but a Twitter account appeared, tweeting under the guise of @Hoyts_Mouse.

The ‘mouse’ started tweeting mid-afternoon on 2 April, and their early activity was soon accompanied by a relative flurry of tweets, questions, and retweets (as with previous datasets featured here, the following analysis does not encompass button retweets, which in this case account for several further @mentions/redissemination of @hoyts_mouse and its comments). Tweets ostensibly from the matinée screenings continued the next day, although the total activity and responses were gradually declining on a day-by-day basis; by Thursday, only once did the number of relevant tweets per hour equal or exceed five.

 

image

The ‘mouse’ was a willing participant in dialogue with its followers, though (indeed, the number of followers initially grew rapidly after Tuesday afternoon on: 114 followers by Tuesday evening, 255 on Wednesday evening, 283 on Thursday evening). Its early tweets covered a variety of topics, including favourite films (from rodent-oriented favourites such as Stuart Little and Ratatouille, to less-expected fare such as Die Hard), the food at Carousel, the films currently screening (with a new Die Hard film and G.I. Joe II out, Bruce Willis was mentioned several times), and its own news media appearances. Several of its comments were also responses to Twitter users pointing out that its profile picture did not quite match the more straggly look of the rodent in the videos and photos uploaded by Carousel patrons.

Among the users tweeting at – and receiving replies from – the Hoyts Mouse account are several Perth journalists, although these are not the only participants in these tweeted discussions. In the visualisation below, I’ve taken the network of @mentions and @replies from the hoyts_mouse dataset, to establish the connections through tweets between the parody account and other Twitter users. In particular, this visualisation helps to show if these tweets form part of long discussion chains involving several users, or if the conversations taking place are primarily dialogues (or, indeed, unreciprocated mentions). As usual for my visualisations, node size here is based on mentions received; the more times a node is mentioned by another, the larger it appears on the map.

hoyts_mouse4

I’ve left most of the node labels off this visualisation, since it gets quite crowded with them on, but what is quite apparent from this network is the type of communication taking place around the @hoyts_mouse account: conversations are generally involving one or two other accounts, with only a few threads developing into multi-party discussions (see the cluster at the top-right, for example, which features tweets mentioning @todaytonight and @wtvbuzz, for example). While the Hoyts Mouse does respond to other users, these replies generally remain directed at individual users (and any further replies might not introduce other users to the conversation). Of course, this does not mean that there are not other tweets responding to those collected in the dataset; however, if they did not include a mention of @hoyts_mouse, then they are not featured here.

What these initial patterns suggest then is that @hoyts_mouse is representative of the short-lived, topical parody accounts, relying on a particular event or story for its creation, than the longer-running and less context-dependent type of account mentioned earlier. Of course, this may yet change, depending on the time and inspiration that the author behind the account has – and the work of the account’s followers in encouraging further activity. With Hoyts now required to address the rodent problem in the Carousel cinema, it will also be interesting to see what effect this has on the account’s lifespan – whether the eradication of the rodent problem will mean the end of the Hoyts Mouse, or if it inspires further rogue posting (and a possible move towards more discussion of current events outside of the context of the initial story). In this latter case, the account then might transition from one form of spoof to another, potentially adding another voice to the various satirical takes on Perth on social media.

 

WA Votes 2013–Election Day (votes and sausages)

By On 2013/04/02 · Leave a Comment

This post is also available at the QUT Social Media Research Group site. It was written in early April 2013, but I’ve been waiting for the site to go live (which it kind of is and isn’t at the moment) before publishing here too.

This post continues my analysis of Twitter data concerning the WA state election, held on 9 March 2013: see this post for an introduction to this project. Previous posts have looked at election-related hashtags, candidates’ posting patterns, and the networks formed between candidates through @mentions, replies, and retweets.

It’s now been over three weeks since the WA state election, which was ultimately won convincingly by the Liberal Party, with a continued swing towards them following their victory in the 2008 state election. Although the analytical posts have been absent since election day, the research has been continuing behind the scenes – as mentioned previously, Axel Bruns and I will be co-authoring a paper about the election as seen from Twitter for the ANZCA conference, to be held in Fremantle in July.

For now, then, a brief return to the datasets to have a look at what happened on election day. In this post, I’m focusing primarily on the #wavotes hashtag, as the main marker used during the second half of the election campaign – later analysis will return to the other campaign-related hashtags and keywords tracked, though. As always with these posts, the usual caveats apply here: the users and views collected are not representative of the general electorate, and not all relevant tweets will have been archived here. The yourTwapperkeeper data capture does not have access to the unlimited stream of tweets from the Twitter API, and in particular, ‘button’ retweets are not always archived. In addition, the network maps do not imply any agreement or support of the users connected (or that the context is an election-specific tweet), but merely that a connection exists through one or more tweets.

Activity using the #wavotes hashtag saw a gradual increase as the election approached, with a previous spike noted during the televised leaders’ debate in February. This is in keeping with other election campaigns covered on Twitter, from Queensland to Scandinavia; similarly, election day itself led to the largest spike in activity, following patterns established in other studies of elections on Twitter. The graph below shows the intensity of the election day spike in #wavotes tweets – jumping from under 2,000 tweets the day before to over 10,000 tweets collected on 9 March itself (see too the similarly dramatic drop-off the day after the election).

 

#wavotes, posts per day around election day

#wavotes, posts per day around election day; vertical axis = total tweets, horizontal axis = day

If we then look at election day itself, though, tweeting #wavotes does not follow a regular pattern across the entire day. Instead, breaking the day down into hour-long sections, we see that there is somewhat uniform levels of tweeting during the early polling period itself, from 8am onwards, and then increased activity as the close of polls and announcement of results approaches. The results and analysis phase of election day accounts for the main #wavotes coverage here; this is, however, fairly short-lived (decreasing activity after 9pm), as the overall election result itself was called early on the night. Rather than hours of speculation about the possible results, the Liberal Party’s success was evident soon into the count, and commentary instead turned to the extent of this victory.

#wavotes, tweets per hour on election day, distinguishing between types of tweet

#wavotes, tweets per hour on election day, distinguishing between types of tweet; vertical axis = total tweets, horizontal axis = hour

 

Not only did election day provoke increased #wavotes activity, but these tweets were also from a greater range of Twitter users. Earlier in the election period, the #wavotes network was focused more directly on the candidates and parties involved in the campaign (with some participation by said candidates, although not to the same extent as in the candidate-specific maps featured in previous posts). However, the heightened #wavotes activity on election day (and the days immediately surrounding it) also generates a network that, in a sense, is more generally covering the election; as the results come in, observers from WA and interstate comment on the trends without necessarily discussing (or needing to discuss) the individual electorates involved. Below is a section of the network map from the #wavotes tweets (based on tweets published between 8 and 10 March, filtered to only include nodes with a degree range of ten or more). As with previous maps, I’ve coloured nodes based on their party affiliation, although only for state candidates. However, the activity by these accounts, with the primary exception of the WA Liberals Twitter account, was relatively minor on election day itself.

To show more clearly the presence of these accounts within the map, then, I’ve set the edge colour to reflect the target, rather than the source, of the edge: a red edge is then directed at a WA Labor account, not necessarily from it (e.g. a tweet mentioning @MarkMcGowanMP). What is then apparent is that, while there was some party-related activity, a lot of the #wavotes connections link users who were not directly involved in the campaign (this includes federal politicians, such as Julia Gillard, Tony Abbott, and Julie Bishop (who was a commentator for the ABC on election night)). Instead, further clusters of users are identifiable within the network; one of these I have highlighted below in yellow, which refers to accounts of ABC journalists, and ABC radio and television stations and programmes. Given the timing of this activity around election night, the centrality of analyst Antony Green is not unexpected here; however, the mass of yellow around Green highlights not just that the ABC was a primary source of information for election coverage, but also the rebroadcasting role of many of these ABC accounts – Green’s own tweets were subsequently retweeted by other ABC accounts, and spread further by other users following these additional, regional or town-specific accounts.

#wavotes, election day @mention/retweet network map (detail)

#wavotes, election day @mention/retweet network map (detail)

 

While the extent of the swing to the Liberal Party was the story of the election night (if the success of the Liberals overall was not unexpected, the results and close races in individual electorates such as Perth and Midland were), during election day there was concerted Twitter activity on another aspect of the voting process; the democratic act of going to a local polling place, voting, and then partaking in a sausage sizzle. The night before election day, Twitter user @bogurk floated the idea of a central hashtag – #democracysausage – in order to collate information about polling places with or without food options. During election day, as users tweeted the sausage sizzle status of their local schools, churches, and community centres, the relevant details were then used to generate and update a #democracysausage Google Map.

Unsurprisingly, the #democracysausage activity follows a different pattern to #wavotes on 9 March; where the latter hashtag spiked during the coverage of the results, #democracysausage – as an election day activity – peaks around midday (lunchtime!), and tails off as booths close, as seen in the graph below. In particular, the number of original tweets decreases after mid-afternoon – while booths remained open until 6pm, the possibility of sausage sizzles or the need to provide information about previously-tweeted sites may decrease in the last few hours of voting. However, it should also be noted that this was an experimental hashtag with a few hundred relevant tweets in the archive, which was only suggested the night before election day and with further aspects developed over the course of the polling day (many of the retweets here, then, were spreading the initial idea behind the hashtag). The people behind the hashtag and map have expressed their keenness in repeating the project at the national scale for the federal election, which will then provide the opportunity to see whether or not these patterns are replicated. [For more, see the @DemSausage Twitter account.]

#democracysausage, election day tweets per hour, distinguishing between types of tweet

#democracysausage, election day tweets per hour, distinguishing between types of tweet; vertical axis=total tweets; horizontal axis=hour

 

new article, upcoming presentations: Australian politics on Twitter

By On 2013/03/27 · Leave a Comment

It’s been a bit quiet here since the WA election on 9 March, although the work on that project has continued – I’ve been saving the final analysis for conference papers (see below), but will provide a round-up of the election on Twitter once that is all done. With the election out of the way, too, other work has taken priority, but the Easter break should give me some more writing time.

While the WA election work continues, though, the journal article that Axel Bruns and I wrote about Twitter and last year’s Queensland state election has just been published online by Information, Communication & Society, ahead of a themed issue around social media and elections (following on from the similarly-themed ECREA panel last year):

Bruns, A., & Highfield, T. (2013). Political Networks on Twitter: Tweeting the Queensland state election. Information, Communication & Society. doi:10.1080/1369118X.2013.782328

Full citation details for this paper will follow when the print version of the journal appears; on that note, I’m also happy to announce that the Information, Communication & Society AOIR special issue details have been confirmed – this edition includes the article with Axel and Stephen Harrington about Eurovision, social media and television audiences, and fandom (along with many awesome papers from last year’s conference – do check them out!). The updated citation for this article is:

Highfield, T., Harrington, S., & Bruns, A. (2013). Twitter as a Technology for Audiencing and Fandom: The #Eurovision phenomenon. Information, Communication & Society, 16(3), 315-339. doi:10.1080/1369118X.2012.756053

The work we carried out on the Queensland election also informs some upcoming papers, both solo and co-authored, that I’ll be presenting in a few months’ time (and, in the long-term, will also be a key foundation for our tracking of the federal election planned for September).

In June, I’m off to Paris to present at an ECR day as part of the “Online political participation and its critics” symposium run by the electronic democracy (DEL) network. My paper there is on “Twitter and Australian political debates“, and builds on the preliminary work I presented, along with Axel and Stephen, at IR13 last year.

A few weeks later, at the start of July, I’ll be back in WA, in Fremantle to present at this year’s ANZCA conference. There, I’ll present a paper co-authored with Axel, entitled “#wavotes: Tracking candidates’ use of social media in the 2013 Western Australian state election“. This paper takes the data collected during the WA election campaign and expands on the initial analyses featured here to more thoroughly explore how Twitter was used in the lead-up to the vote. Covering both the Queensland and WA cases also provides an opportunity for comparing similar types of campaign – state-level politics – which will then contrast with the federal coverage later this year.

There are more projects in progress at the moment, too; right now, my Curtin colleague Sky Croeser is in Tunisia for the World Social Forum for the second stage of our “Mapping Movements” research, and, as with our earlier research into Occupy Oakland, I’m supporting her on-site studies by tracking relevant tweets and starting web crawls from central websites (rediscovering IssueCrawler in the process). More details on this, and other projects, to come later!

 

WA Votes 2013–last week of the Twitter campaign

By On 2013/03/09 · Leave a Comment

Over the next couple of months, I will be collecting and analysing Twitter data (technology permitting) around the WA state election, to be held on 9 March 2013: see this post for an introduction to this project.

Today, Western Australia goes to the polls – booths are open, and will remain so until 6pm today, so if you are enrolled to vote please do go and number boxes! More details on polling places, procedures, and times are at the WAEC’s WA Elections website.

I haven’t been able to post as regularly during the campaign as I initially intended when I started this project, and today is just going to be a brief update – the data collection has been continuing, though, and with a bit more time in the next couple of weeks I’ll post some more analysis of the election data, including election day and the aftermath.

For today, I’m just going to provide the network map of the candidates over the first week of March (1-7 March), and some candidate activity statistics up to the opening of the polling booths at 8am today. These patterns give us some indication, as usual, of which candidates are being mentioned, replied to, and retweeted by other candidates during the campaign, and which candidates are using Twitter the most.

Previous candidate maps have noted the presence of partisan clusters, especially around the Greens and the Nationals (the Liberals have a more limited representation on Twitter, and the ALP dominate in terms of numbers). Looking at the mentions of candidates during the week 1-7 March, these patterns generally continue, although with some exceptions:

 

wavotes2013_polandpartyonly_firstweekofmarch

Here, size of nodes is based on in-degree – the number of times an account is mentioned by other accounts – and colour is based on party affiliation. What is clearly apparent again is a Nationals cluster at a distance to the rest of the network, but with some strong inter-linking between Nationals candidates – apart from Vincent Catania, who appears within the ALP-dominated section through mentions by his rival candidate Jennifer Shelton (both are running in the North West electorate).

Among the ALP candidates are also several independent candidates; the previously discussed Greg Ross, running in Kalamunda, and also Michael Tucack (@avibrantcity) running in the North Metropolitan division for the Legislative Council – I hadn’t been tracking Tucack’s account during the campaign, but have added it to the list just before election day itself (as with a few other independent/minor party categories, including the minor media sensation that is Family First’s Henry Heng). Although the ALP candidates again form a densely inter-linked section of the network, there are several cases where individual candidates from the ALP and the Greens mentioned one another, drawing the two clusters closer together (ALP and Greens candidates still primarily mention their fellow party members, but there are some cross-party mentions).

If we turn now to the tweeting activity over the last month – 6 February to 9 March – we can identify who has been most prolific, in terms of total tweets (including retweets of other users) during the campaign (this period goes up to 8am today, so does not include the election day polling period – that’s a topic for a later post). Regular readers will be unsurprised to see the ALP’s Ken Travers at the top of the list, with 740 tweets over this period, and he is followed by other accounts whose high activity has been a recurring pattern; the Greens’ Cameron Poustie and George Crisp (and the party account), the ALP’s John Hyde and Paul Papalia, and Greg Ross.

Name

User name

Party

Tweets (6 Feb-9 March)

Ken Travers

KenTraversMLC

ALP

740

Cameron Poustie

CameronPoustie

Greens

483

The Greens (party)

TheGreensWA

Greens

456

George Crisp

DrGCrisp

Greens

399

Greg Ross

GregWRoss

Independent

397

Paul Papalia

papsMLA

ALP

284

John Hyde

JohnHydeMLA

ALP

272

Joe Francis

JandakotJoe

Liberal

179

Liberals (party)

LiberalsWA

Liberal

159

Barry Urban

darlingrange

ALP

131

What’s also interesting, though, is to evaluate whether this activity is a regular pattern for the candidates themselves, or if the election represents increased use of Twitter. Treating the number of tweets published between 6 February and 9 March as a percentage of each account’s total output to this point, we see that many of the prolific accounts drop off the list – their activity is in keeping with pre-election output, or they have been active on Twitter for a long time. Again, unsurprisingly, we get accounts set up for new candidates, and launched for the election, as the majority of entries in this table. Indeed, the ALP’s Sharon Webb, who has not previously been mentioned here, tops the list with over 90% of her tweets coming during the campaign. However, it should also be noted that most of her tweets are automatic updates from Facebook of photo uploads from ALP fundraisers, rather than replies to other users and comments from the candidate herself.

Name

User name

Party

Tweets (6 Feb-9 March)

% tweets (to 9 March)

Sharon Webb

SharonWebbLabor

ALP

39

90.7

Kevin Morgan

KMCottesloe

Independent

33

64.7

The Greens (party)

TheGreensWA

Greens

456

53.5

Greg Ross

GregWRoss

Independent

397

53.4

Klara Andric

KlaraAndric

ALP

113

50.9

Rob Phillips

GreensWanneroo

Greens

31

42.5

Joe Francis

JandakotJoe

Liberal

179

41.8

Sarah Newbold

sarahnewbold

Greens

122

39.9

Jennifer Shelton

JSheltonALP

ALP

109

39.8

Andrew Sullivan

Sullivan4Freo

Greens

108

38.0

Barry Urban

darlingrange

ALP

131

37.5

While many of these entries are new candidates, two accounts do stand out here: the Greens’ party account, which with over 50% of tweets coming during the campaign has had a clear increase in social media activity as part of its election strategy, and the Liberal’s Joe Francis. In this case, Francis was rather quiet earlier on in the year, but (maybe by default) became a Liberal spokesperson on Twitter, possibly because of the absence of other major Liberal accounts – Francis is a current member of Parliament and has been on Twitter for a while, but with over 40% of his output coming over the last month or so, he has increased his tweeting to respond to, or provoke comments from, other candidates and parties. After the election, I’ll have a more extensive look at individual accounts and their approaches to tweeting during the campaign – alongside many other posts around hashtags, keywords, and links, which I’ll try to write up in the next few weeks!