NFG-WInDIATA11 Program

DIATA11 Program

Düsseldorf Workshop on Interdisciplinary Approaches to Twitter Analysis (#diata11)

14.-15. September, 2011, Düsseldorf, Germany

University and State Library Düsseldorf, Building 24.41, large lecture room

Convened by the Junior Researchers Group "Science and the Internet" (NFG-WIn) and the CCI Media Ecologies and Methodological Innovation Project under the ATN-DAAD Australia Germany Joint Research Cooperation Scheme.

Download the program as a PDF

Wednesday, September 14, 2011

Time

Session

10.00-12.00

Pre-Workshop Coding Session
Building 23.21, Room 00.85

13:00-13:30

Opening Session

Welcome message from Prof. Axel Buchner, Heinrich-Heine-University's Vice President for Academic Management and International Relations

 

Opening remarks by Katrin Weller and Cornelius Puschmann (slides)

13:30-14:00

Keynote

Mapping Online Publics on Twitter (slides)

Axel Bruns, Jean Burgess, and Stephen Harrington

14:00-14:30

Coffee Break

14:30-16:00

Session 1: Metrics and Measurement

Twitter and the New Regime of Audience Measurement (slides)

Jan Schmidt

 

How to Measure Opinion Leadership on Twitter (slides)
Katrin Jungnickel

 

Twitter as a Data Source for Official Statistics: Usability and Classification

Piet Daas, Marko Roos, and Joyce Neroni

16:00-16:30

Coffee Break

16:30-18:00

Session 2: Politics and eParticipation
Mapping the Austrian Political Twittersphere
Julian Ausserhofer

 

Twitter Use During the 2010 Swedish Election – and Beyond (slides)

Anders Larsson

 

One Tweet at a Time: Mapping Political Campaigns Through Social Media Data

Andreas Jungherr and Pascal Jürgens

19:30

Dinner at Füchschen Brewery

 

Thursday, September 15, 2011

Time

Session

09:00-10:30

Session 3: The Impact of Twitter on Scholarship and Journalism

Public Archaeology and Twitter – Widening the Debate in 140 Characters or Less (cancelled)

Lorna Richardson

 

The Use of Twitter in Professional Journalism – Results of a Newsroom Survey in Germany

Christoph Neuberger

 

Evolution of News: Dynamics between Social and Mass Media (slides)

Axel Maireder

10:30-11:00

Coffee Break

11:00-12:30

Session 4: Semantics and Sentiment

Detecting Semantic Entities from User Tweets (slides)

Smitashree Choudhury

Information Propagation and Extraction from Twitter (slides)

Matteo Magnani

Customer Communication in Twitter (slides)

Stefan Stieglitz, Nina Krüger, and Linh Dang-Xuan

12:30-13:30

Lunch at Campus Vita

13:30-15:00

Session 5: Communities and Conversations

Uncovering Subgroups Among the Masses: A Group Informatics Approach to Twitter (cancelled)

Christopher Mascaro

 

Twitter, Television and Fußball (slides)

Stephen Harrington

 

Twitter as a Discourse Universe: a Mixed-Methods Approach Towards the Analysis of Semiotics and Speech Acts in Tweets (slides)

Mark Dang-Anh and Jessica Einspänner

 

Beyond the Stars: Interpreting Discourse Cohesion in Twitter as an Indicator of Community (slides)

Cornelius Puschmann

15:00-15:30

Coffee Break

15:30-16:30

Discussion

 

Abstracts

Mapping the Austrian Political Twittersphere

Julian Ausserhofer
Department of Journalism and Public Relations
Graz University of Applied Sciences

Unlike in the United States, where Lady Gaga and Justin Bieber rank among the most followed accounts, in Austria, celebrities are no 'Twitter stars': 14 of the top 20 Austrian Twitter accounts followed by Austrians are maintained by journalists (Social Media Radar Austria 2011; http://socialmediaradar.at/). The first account not related to journalism is placed eighth. Furthermore, politicians have increasingly discovered Twitter as a medium of political dialogue. This has led to the development of a specialized public sphere where politicians, journalists, experts and lobbyists discuss certain topics. The conversations obviously shape (and are shaped by) the agenda of the (highly concentrated) Austrian media landscape. This research project in development to be presented initially at #diata11 analyzes the political Austrian Twittersphere and examines the mutual dependencies between traditional media and Twitter.

Since June 2011 Axel Maireder, Axel Kittenberger and I have been looking for Austria’s political twitterers by compiling lists and searching for the mention of certain keywords and leading politicians with Twapperkeeper. We are currently in the process of manually categorizing the twitterers of the Austrian political sphere by their professions. Later this year we will record their whole twitter output for two weeks. To identify the influencing actors we will perform a quantitative network analysis based on their connections and conversations using Gephi (and other tools). We additionally plan to cluster and compare the domestic political agenda on Twitter and on traditional media in order to gain insight on who influences whom and with what effect. At DIATA '11 I would like to discuss our methodical setting and present first findings on the Austrian Twittersphere which has not been mapped before.

This research project was developed out of Axel Bruns' seminar "Methoden für die datengestützte Studie themenfokussierter Öffentlichkeiten im Netz" which he had held earlier this year at the Institute for Media and Communication Studies, University of Vienna.

 

Mapping Online Publics on Twitter

Axel Bruns, Jean Burgess, and Stephen Harrington
ARC Centre of Excellence for Creative Industries and Innovation
Queensland University of Technology

This paper reports on methodological developments and research outcomes from the Mapping Online Publics project at QUT. Our research focus has been on understanding the activities of Twitter users around hashtags, especially in the context of ‘acute events’ ranging from natural disasters (the Queensland floods, the earthquakes in Christchurch and Japan) through political crises (leadership challenges and elections) to cultural events (the British royal wedding, major TV events). In the course of our research, we have developed a rich collection of tools for processing, analysing, and visualising large Twitter datasets.

In addition to this focus on hashtag-based interactions, we are also establishing the first comprehensive map of follower/followee relationships in the Australian Twittersphere, in order to understand the underlying structure of the social network of Australian Twitter users and to identify the clusters of shared interest and focus which may exist within it. In the course of this long-term project, we are also working towards a more reliable approximation of the total size of the Australian Twittersphere. This paper will present a progress report on this mapping exercise.

 

Detecting Semantic Entitles from User Tweets

Smitashree Choudhury
Digital Enterprise Research Institute (DERI)
National University of Ireland, Galway

Popularity of tweets as part of the online discourse skyrocketed in recent years when every event of any importance are being reported and discussed by multiple users with short messages. Constrained by the number of characters , tweets are often aimed to convey the message at the cost of the grammar rules. Extracting knowledge from these content is non- trivial and an emerging challenge for the computer assisted human language understanding research community.

As part of this study, we took two different domains (sports, conference) of tweet recorded during multiple live events to detect various topics and semantic entities discussed by users. The study explored various approaches (natural language and statistical methods) to detect these topics and interesting micro-events. Experiment showed recall increases with combinations of multiple approaches compared to a single technique. Some domain like sports depends on domain knowledge to detect entities and events more than conference domain. Application of this study is aimed to achieve a timed stamped annotation of the video recorded during the live event.

 

Twitter as a data source for official statistics: usability and classification

Piet Daas, Marko Roos, and Joyce Neroni
Centraal Bureau voor de Statistiek (Dutch National Office of Statistics), Den Haag

The social medium Twitter is used by more and more people to discuss topics of interest, exchange information, and/or contact friends and family. At Statistics Netherlands we investigated the content and usability of the information exchanged by Dutch Twitter users from an official statistics point of view. First, various ways of collecting Twitter text messages were compared. User oriented data collection was found the best approach. Next, the usability and classification of the topics discussed in Dutch Twitter messages was investigated. The topics discussed in Twitter messages were compared to the various themes Statistics Netherlands is interested in. Automatic and manual classification results revealed that close to 50% of the Dutch Twitter messages collected could be allocated to themes of interest. Classification of these messages to the various themes proved challenging. This was, amongst others, caused by the distorting effect of the non-relevant Twitter messages. We therefore also looked at ways to filter out the relevant messages prior to analysis. On overview of the results obtained so far and ideas for future research will be presented.

 

Twitter as a discourse universe: a mixed-methods approach towards the analysis of semiotics and speech acts in tweets

Mark Dang-Anh and Jessica Einspänner
Department of Media Studies
University of Bonn

Processes of mediatization are considered as a societal meta-processes (Krotz 2007). Social media services such as Twitter can, within this conceptual frame, be thought of as transmedial discourse universes which open up new ways of political participation in the notion of deliberative democracy. Tweets are constituted on the basis of a specific semiotic system that not only implies the use of digital script – and thereby the creation of speech acts – but also facilitates a small but complex range of acts of reference. Four operators of communication (Kommunikationsoperatoren) can be identified in this regard: @, #, http:// and Retweets (Thimm/Dang-Anh/Einspänner 2011). About 26.000 Tweets by politicians, party-accounts and media-accounts were collected during the 2011 federal state elections in Baden-Wuerttemberg. In our presentation, we rely on a selective sample of this dataset. Tweets from this sample are being analysed in a combined qualitative and quantitative content analysis (Mayring 2007; Welker/Wünsch 2010) in order to systematically detect interrelations between operators of communication and speech acts. Furthermore, fundamental issues concerning the collection and evaluation of data in Twitter will be discussed.

 

One tweet at a time: Mapping political campaigns through social media data

Andreas Jungherr
Research Associate, Chair of Political Sociology
University of Bamberg

Pascal Jürgens
Department of Mass Communication
University of Mainz

Data collected from internet platforms provide researchers with new insights into human behavior online. As stated by the Computational Social Science (Lazer et al. 2009) and Digital Methods (Rogers 2009) research programs these data also offer more general insights. Recently the microblogging service Twitter has become a popular tool for users to comment on political campaigns. The automatic collection of these messages is possible since users usually include a campaign specific #keyphrase. Users not only comment on politicians or parties but also post links to news articles and blog posts. This allows researchers to map campaigns: The frequency of twitter messages containing a campaign specific #keyphrase allows researchers to map the intensity of a given campaign over time; the frequency of words contained in campaign specific twitter messages allows researchers to map topics dominating specific campaigns during specific time frames; links to news content allow researchers to understand what elements of the media environment dominate the public discourse. We demonstrate this potential by analyzing twitter messages commenting on one of the most controversial recent public discussions in Germany, Stuttgart 21. The basis for this analysis are twitter messages by the 80.000 most active German twitter users collected through the Twitter API during 2009. Using all messages containing the keyphrase #21 we map the development of the public discussion with regard to message frequency over time, dominating keywords in campaign specific messages, and media content prominently linked in campaign specific messages.

 

How to measure opinion leadership on Twitter

Katrin Jungnickel
Department of Public Relations and Technical Communication
Technische Universität Ilmenau

With the discovery of the two-step flow of communication (later: multi-step flow) in the 1940s, the concept of opinion leadership and the two main functions of opinion leaders – (1) relaying information from professional communicators (media, politicians, organizations) to the public (transmission) and (2) influencing public opinion, attitudes and behavior (persuasion) – became crucial for researchers of political and organizational communication. In online social networks, the impact of opinion leaders acting as micro-celebrities, gatekeepers for information, and influencers potentially increases because their network size has expanded considerably with some having thousands of followers and friends. Furthermore, the amount of data provided by online social networks allows researchers to develop new methods of measuring opinion leadership in addition to traditional scales (e.g. King & Summers 1970, Noelle-Neumann 1983) relying on self-assessment, or reputational approaches based on the nomination of opinion leaders by others. On Twitter, different strategies have already been tested with calculating the amount of retweets and reposts reported to be the most effective in identifying intermediaries that transmit information to others (e.g. Kwak et al. 2010, Cha et al. 2010). But those technical approaches often lack a theoretical framework and focus rather on the transmitting function of opinion leaders than on their role as influencers and persuaders. The presentation aims to link the traditional concept of opinion leadership to the new online gatekeepers and micro-celebrities, and will discuss possibilities for combining traditional and new methods to identify opinion leaders in online social networks, especially on Twitter.

 

Twitter use during the 2010 Swedish election – and beyond

Anders Larsson
Department of Informatics and Media
Uppsala University

Since its launch in 2006, Twitter use has evolved and is increasingly used in a variety of contexts. One such context is online politics or e-democracy. Although many of the initial hopes for e-democracy appear to have gone largely unfulfilled, the successful employment of the Internet during the 2008 US presidential campaign has again raised voices claiming that the Internet, and particularly social media applications like Twitter, provides interesting opportunities for online campaigning and deliberation. My presentation in Düsseldorf will focus on an ongoing comparative study involving researchers in the scandinavian countries, dealing with how political use of Twitter is fashioned in the three countries of Denmark, Norway and Sweden. Specifically, I will present findings pertaining to Twitter use during the 2010 Swedish election. I would also like to discuss some of the ethical challenges that we have come across during this work.

 

Evolution of news: dynamics between social and mass media

Axel Maireder
Department of Communication
University of Vienna

In the past, diffusion research was mostly done through telephone interviews. Today, social media not only allows us to trace the diffusion of news but the evolution of a news story through content analysis. We set up and tested a research design to analyse the cross-medial diffusion and evolution of news in different, clear-cut cases, asking for how the dynamics are shaped by different types of media. We also ask for the differences in the way news media reports on a story and the way social media users talk about it.

We code publishing time and content aspects (distinct pieces of information of central relevance) of a wide array of available texts including news stories in online newspapers, blog entries and tweets, that all deal with a certain topic. After homogenizing the material structure and chronologically ordering it, we start by selecting the first item and defining the main aspects included in its content. For each of the following texts, we code whether those aspects are included or not. Each time a new aspect, a new piece of relevant information is added in a text, the code sheet is extended respectively. Our goal is to trace the evolution of a story according to the aspects people write about.

At the time of writing this abstract, the development of our methodology is more-or-less finished, but I am very much looking forward to present and discuss the setting at the Workshop. I hope I will also be able to present first results, the coding of the first topic is planned to be done in August.

 

Information propagation and extraction from Twitter

Matteo Magnani
Department of Computer Science
University of Bologna

In this talk I will present some results of the #sigsna research project concerning the extraction of information from Twitter and the study of information propagation patterns. In the first part of the presentation I will introduce the topic of Conversation Retrieval, a methodology to filter Twitter conversations based on metrics computed on whole conversational interactions and not just single tweets. The importance of the participants, the timeliness, length and density of conversations are used together with traditional keyword-queries to isolate relevant information. An interesting research hypothesis validated by preliminary experimental studies shows that this information extraction process is more appreciated by users than traditional search tasks and that the tuning of the conversation ranking parameters may be influenced by the cultural context of the users. Then I will present a case study concerning the live coverage of a global event (the rescue of 33 Chilean miners after a mining accident) on Twitter, showing the role of traditional mass media accounts in this process and highlighting the fact that local nation-based identities are still important and can be observed as clusters of interconnected conversations.

 

Uncovering subgroups among the masses: A Group Informatics Approach to Twitter

Christopher Mascaro
College of Information Science and Technology (iSchool)
Drexel University

The advent of technologically mediated social networking tools such as Facebook and Twitter have created a need to develop new methods for collecting, analyzing and interpreting the interactions that occur within each specific technology. The Group Informatics Lab at Drexel University is currently developing methodological approaches that take into account the different manifestations of technologically mediated group behavior in these specific technologies. These methods attempt to understand how individuals interact and form groups in a technologically mediated environment and the ways that these actions can be analyzed utilizing existing and newly developed approaches.

In this talk we will first offer an overview of our existing methodological approaches for analyzing social media technologies such as Facebook and discussion forums. Second, we will present previous findings from a variety of contexts to include: political discourse, software engineering and disaster relief. We will use the latter half of the talk to share early results of our attempts to identify emergent groups in Twitter through the utilization of the rich network ties and technological affordances of the technology. We will conclude the talk by detailing our research agenda and how we intend to further develop and apply our analytical methodologies to Twitter.

 

The Use of Twitter in Professional Journalism – Results of a Newsroom Survey in Germany

Christoph Neuberger
Department of Mass Communication
University of Munich

Twitter is a multifunctional tool for public communication which allows different types of users to disseminate information. How is Twitter adopted in professional journalism? This was the main research question of a newsroom survey which was conducted in May and June 2010. Editors-in-chief from online newsrooms were asked about the manner of use of Twitter and other social media. 70 editors participated (response rate: 45 percent). Results show that almost all media organizations used Twitter to attract readers and for investigation. About two third of participants said they used Twitter for interaction with users and also for live coverage and breaking news. Twitter is part of the new news ecosystem in which media and audience are mutually connected. Although Twitter offers a lot of options for professional journalism, it is only one of many channels on the Internet. Journalism is now in a learning process to find the special ways of use for the microblogging service. Especially the brevity of tweets limits its capabilities. Another restriction is the short reach of Twitter in Germany: Only 3 percent of Internet users ever used Twitter at least seldom in 2010.

 

Beyond the stars: Interpreting discourse cohesion in Twitter as an indicator of community.

Cornelius Puschmann
Department of English
University of Düsseldorf

Twitter is widely characterized as a channel for distributing information in a one-to-many fashion (Kwak et al. 2010). While many of the plattform's sociotechnical properties favor this approach, it is not uncontested by other modes of use. Honeycut & Herring (2009) point to an interesting example of such unconventional use in linguistic terms, namely conducting complex, multi-turn conversations using the @-sign.Their analysis suggests that Twitter users frequently engage in such exchanges, repurposing Twitter as a tool for conversation, rather than broadcasting.

My talk will explore @-conversations among a network of (mostly German) gaming enthusiasts and pay special attention to how the cohesion of conversations -- in spite of Twitter's "anti-conversational" properties -- mirrors the social cohesion of the community. I will proceed to argue that follower/followee metrics carry mostly political relevance (in sociocommunicative terms) and are misleading as reliable indicators of community. I will also counterpoint the conversational behavior of "stars" (i.e. very popular Twitter users with a large number of followers) with that of casual users.

 

Public archaeology and Twitter - widening the debate in 140 characters or less

Lorna Richardson
Department of Information Studies
University College London

In 2011, my PhD research into public archaeology and internet technologies has concentrated on the use of Twitter within the archaeological sector, both as a medium for networking and information-sharing within the profession, and as a platform for engagement with the general public about archaeological subjects. As part of my research, 165 archaeologists completed a 27-question survey about their use of Twitter. Concurrently, a Twitter-born project, the 'Day of Archaeology' was created via a Twitter conversation between myself and a collegue in Wales, and on the 29th July, over 400 archaeologists from around the world took part in a blogging project propagated almost solely through Twitter. This experiment on community-building within Twitter is currently being written up.

 

Twitter and the new regime of audience measurement

Jan-Hinrik Schmidt
Hans-Bredow-Institute for Media Research

The social web is bringing about new transformations of the public sphere. Twitter, in particular, is introducing innovative mechanisms of distributing information among articulated social networks of "followers" and "followees", expanding the reach of professional media as well as contributing to new issue publics. The presentation will focus on the consequences of these changes for the "manufacturing of audience", taking Twitter as an integral part of a new regime of audience measurement that is emerging in social media.

For example, Twitter is making visible (and measurable)

  • the "follow-up communication", both within the audience and among other journalistic actors;
  • the structures and centrality-periphery-patterns of the networked audience;
  • the reach and assessment of de-bundled "microcontent";
  • the fluidity and viral spread of this microcontent, when re-bundled within streams and feeds.

The presentation will describe and discuss these facets of twitter-based networked publics and how they are managed through practices of measuring and monitoring the audience. In addition, it will outline some structural aspects of the new regime of audience measurement, including quality aspects of various indicators, the organizational structure of measurement and monitoring, as well as emerging issues of privacy and data protection.

 

Customer Communication in Twitter

Stefan Stieglitz, Nina Krüger, and Linh Dang-Xuan
Department of Business Informatics
University of Münster

Since 2006, microblogging has become a new and widely used concept for communication in the Internet. Twitter, as one of the first and most popular microblogging providers, counts 106 million users, according to their own statements. On average 140 million Tweets (short messages with up to 140 signs) are posted each day (Twitter 2011). In contrast to social networking sites (SNS) such as Facebook, the range of communication in Twitter is not limited to a specific group (e.g. friends). Each Internet user can start discussions, follow the communication of others, and participate into debates. To make this communication more efficiently, Twitter users adapted different ways of marking their contributions (postings), e.g. as an answer (retweet), direct message or mention. This new way of public communication affects enterprises, because (1) customers are enabled to share experiences about companies and products frequently and fast among each others, (2) enterprises may adopt social media as a new channel to communicate with customers, and (3) the user-generated content can be tracked and analysed to extract relevant and innovative ideas.

In our research we seek to analyze the dynamics of a specific issue and the sentiments of that discussion. Therefore, we referred to an inductive approach to identify an appropriate issue (crisis) to monitor its related discussion in Twitter for several months. Based on literature of traditional public communication, we expected that the dynamics of this discussion in Twitter show some stages of "buzzing" (continuous communication that is characterized by a low frequency of postings) as well as "peaks", where significantly more contributions are published (Röttger, U. 2002). Also it has not been investigated if power-tweeters (most active participants of discussion) are using sentiment words more frequently than other users. Until now, to our knowledge there is no research addressing the analysis of sentiments in those different stages.

Research questions are:

  • Are crisis-related long-term discussions in social media characterized by stages of buzzing and peaks?
  • Do involved users of a specific issue generate more contributions (postings) in peaks than in buzzing stages?
  • Are discussions in peaks characterized by “stronger” emotions (sentiments) than in buzzing stages?
  • Are messages, which contain sentiment words retweetet more often than messages without sentiment words?
  • Is there a difference between sentiments in Tweets created by power-tweeters (PT) only and the sentiments in Tweets created by all participants of the sample?

Twitter (#cosci12)