New Approaches to Large-Scale Social Media Analytics: Investigating Twitter in Australia

New Approaches to Large-Scale Social Media Analytics: Investigating Twitter in AustraliaProfessor Axel BrunsDigital Media Research CentreQueensland University of TechnologyBrisbane, Australia

a.bruns @ qut.edu.au@snurb_dot_infohttp://mappingonlinepublics.net/

QUT DIGITAL MEDIA RESEARCH CENTRE

The Digital Media Research Centre (DMRC) conducts world-leading research that helps society understand and adapt to the social, cultural and economic transformations associated with digital media technologies, and trains the researchers of tomorrow.

For more, see: http://www.qut.edu.au/research/dmrc

Journalism, Public Communication &

Democracy

Economies, Policies & Regulation

Digital MethodsTechnologies &

Practices in Everyday Life

DIGITAL MEDIA

DMRC PROGRAMMES

RESEARCH PROJECT

• ARC Future Fellowship:– Four-year project– Axel Bruns (FF), Brenda Moon (Postdoc), Felix Münch (PhD1, 2014-2017);

PhD2 (2016-2018)

At the intersection of mainstream, niche, and social media, the processes by which public opinion forms and public debate unfolds are increasingly complex, and poorly understood. This project draws on large datasets and innovative methods to develop a new model of the Australian online public sphere.

• Also supported by ARC LIEF project:– Two-year project (2014/15; QUT, Curtin, Deakin, Swinburne) to develop

comprehensive infrastructure for large-scale social media data analytics

BIG SOCIAL DATA

(http://siliconangle.com/blog/2011/08/09/twitter-unravel-the-mysteries-of-big-data/)

TWITTER: STATE OF THE FIELD

• Twitter research to date:– Abundance of hashtag studies: volumetrics, keywords, networks, …– Some studies profiling samples of the total userbase (e.g. celebrities, politicians)– Some comprehensive (?) tracking of activities around key events and topics– Some egocentric follower network maps, largely small-scale– Almost absent: comprehensive follower network maps, longitudinal userbase development

trajectories, user career patterns from sign-up to listener/celebrity/…

• The political economy of Twitter research:– Twitter API data access is shaped to privilege certain approaches– Research funding is easier to obtain for specific, limited purposes– Longitudinal, ‘big’ data access requires ongoing, substantial funding and infrastructure– Exploratory, data-driven research is difficult to sell to most funding bodies– Also related to divergent resources available to different scholarly disciplines

Most ‘hard data’ Twitter research conducted by Twitter, Inc. and commercial research institutes

(https://blog.twitter.com/2012/dispatch-from-the-denver-debate)

BEYOND HASHTAGS

• Key needs in Twitter research:– Understand how hashtags are situated in a wider communicative ecology on Twitter– Document the day-to-day uses of Twitter, beyond and outside hashtags– Trace the dynamics of Twitter as a platform for everyday quasi-private, interpersonal,

and/or public communication– Track the impact of social and technological changes on these uses

• ad hoc publics, often rapidly forming

and dissolving

macro: #hashtags

• personal publics, accumulating slowly and relatively stable

meso: follower networks

• interpersonalcommunication,

ephemeral

micro:@replies

(Bruns & Moe, 2013)

THE AUSTRALIAN TWITTERSPHERE

• Twitter in Australia:– Strong take-up since 2009– Centred around 25-55 age range, urban, educated, affluent users (but gradually broadening)– Significant role in crisis communication, political communication, audience engagement, …

• Mapping the Twittersphere:– Long-term project to identify all Australian Twitter accounts– First iteration: snowball crawl of follower/followee networks

• Starting with key hashtag populations (#auspol, #spill, …)• Map of ~1m accounts in early 2012

– Second iteration: full crawl of global Twitter ID numberspace through to Sep. 2013 (~870m accounts)

• Filtering by description, location, timezone fields• Focus on identifiably Australian cities, states, timezones and other markers• 2.8 million Australian accounts identified (by Sep. 2013)• Retrieval of their follower/followee lists• Best guess of account location based on timezone, location and description settings

DAILY GROWTH (GLOBAL)

MONTHLY GROWTH (AUSTRALIA)

MAPPING THE AUSTRALIAN USERBASE

• Mapping the Twittersphere:– Filtered to include only accounts with (followers + followees) >= 1000

• 140k accounts, 22.8m follower/followee connections within this group– Mapped using Gephi Force Atlas 2 algorithm (LinLog mode, scaling 0.0001, gravity 0.5)– Qualitative interpretation of network clusters based on high-degree nodes in each cluster– Exploration of key profile statistics (join date, number of tweets, tweeting rate)

• Activity Patterns:– Data gathered on selected tweeting activities (hashtag participation, link sharing, …)– Data filtered for participating accounts included in the 140k most connected users– Data superimposed on underlying network map

• Applications:– Combined analysis of network structures and tweeting activities– Evaluation of user engagement across topics– Comparative benchmarking across activities

Education

Agriculture

Literature

Adelaide / SA

FoodWine

Beer

Parenting

Mums PR

Netizens

Marketing

InvestingReal Estate

Home BusinessSole Traders

Self-Help

HR / Support

Followback

Urban MediaUtilities

Advertising

Business

Fashion

Beauty

ArtsCinema

Journalists

Politics

Hard RightLeftists

News

CyclingTalkback

Music

TVV8s UFC

NRL

AFL

Football

Horse Racing

CricketNRU

Celebrities

Hillsong

Perth

PopMedia

Teen Idols

Cody Simpson

THE AUSTRALIAN TWITTERSPHERE

PROTECTED (~3.5%)

VERIFIED (~1.8%)

ACCOUNTS CREATED IN 2006

ACCOUNTS CREATED IN 2007

ACCOUNTS CREATED IN 2008

ACCOUNTS CREATED IN 2009

ACCOUNTS CREATED IN 2010

ACCOUNTS CREATED IN 2011

ACCOUNTS CREATED IN 2012

ACCOUNTS CREATED IN 2013

NUMBER OF TWEETS

LOCATION

USER ENGAGEMENT PATTERNS

• Datasets included:– Q&A: political talkshow, Australian Broadcasting Corporation – #qanda, qanda (3 Sep. to 7 Oct. 2014)– Big Brother: reality TV, Nine Network – #BBAU, #BBAU9, @BBAU9, #bigbrotherau (3 Sep. to 7 Oct. 2014)– AFL Grand Final: Seven Network – #AFLGF, AFL, HAWvSYD … (27 Sep. 2014, tracked since 26 Sep.)

– ABC News: main public broadcaster – abc.net.au (June 2012 to Sep. 2014)– news.com.au: mainstream news site – news.com.au (June 2012 to Sep. 2014)– Daily Telegraph: tabloid newspaper – dailytelegraph.com.au (June 2012 to Sep. 2014)– The Conversation: scholarly opinion site – theconversation.edu.au/.com (June 2012 to Sep. 2014)

AFL GRAND FINAL

Q&A

BIG BROTHER

AUSTRALIAN TWITTER NEWS INDEX (ATNIX)

• Data gathering:– All tweets containing links to one of ~35 Australian news sites– Includes tweets with shortened links (t.co, bit.ly, etc.)– Continuous data gathering since August 2012 (with some outages)

• Analysis:– Overall development of sharing patterns– Marketshare of specific news organisations– In-depth analysis of day-to-day trends in sharing activity– Correlation with local, regional, and global events

• Similar projects in place for Germany and the Nordic countries

ABC NEWS

NEWS.COM.AU

DAILY TELEGRAPH

THE CONVERSATION

TRACKING AUSTRALIAN TWITTER ACTIVITY

• Next steps:– Tracking 2.8m Australian Twitter users’ public tweets continuously– Comprehensive dataset of Australian Twitter activities– Full coverage beyond predetermined hashtags / keywords– Important resource for rapid response analytics– Valuable dataset for longer-term historical research

– Live tracker now operational, using Sep. 2013 userbase– Need to regather full Australian userbase in 2015

AUSTRALIAN TWEETS PER DAY

LEADING HASHTAGS IN AUSTRALIA

MOST @MENTIONED USERS IN AUSTRALIA

Sydney Siege, 15/16 Dec. 2014

Peak: ~30.1% of tweets

Peak: ~37.1% of tweets

Paris Attacks, 13-15 Nov. 2015

http://mappingonlinepublics.net/@snurb_dot_info

@socialmediaQUT – http://socialmedia.qut.edu.au/ @qutdmrc – https://www.qut.edu.au/research/dmrc

This research is funded by the Australian Research Council through Future Fellowship and LIEF grants FT130100703 and LE140100148.