You are on page 1of 10

Draft version, originally published in: N. Reynolds and M. Turcsányi-Szabó (Eds.

):
KCKS 2010, IFIP AICT 324, pp. 102–113, 2010.

Get Granular on Twitter -


Tweets from a Conference and their Limited Usefulness
for Non-Participants

Martin Ebner1, Herbert Mühlburger1, Sandra Schaffert4, Mandy Schiefner2,


Wolfgang Reinhardt3

1
Social Learning, Computer and Information Services, Graz University of Technology
Graz, Austria
2
Center for University Teaching and Learning, University of Zurich,
Zurich, Switzerland
3
Computer Science Education Group, University of Paderborn
Paderborn, Germany
4
Salzburg Research Forschungsgesellschaft, Salzburg, Austria

Abstract. The use of microblogging applications (especially Twitter) becomes


common in different settings. Nowadays, active conference participants post
messages on microblogging platforms to exchange information fast and in real-
time. Recent research work addressed quantitative analyses in terms of number
of tweets or active Twitter user in a specific time period. In this publication we
look at the content of the contributions. It can be shown that only a few posts
are of interest for non-participants of the corresponding event and that a
meaningful usage of a microblogging application needs carefulness.

Keywords: Twitter, microblogging, conference, analyses

1 Introduction: Usage of Twitter at conference

Twitter is the most popular public microblogging system. After a period of testing this
form of communication and interaction in science [2] and e-learning [1] [12] [13],
microblogging finally caught up with the scientific community: Some early adopters
use it to share short notices about their work or to comment other work, they use it to
communicate and last but not least, they use it at conferences. In fact, many
researchers report that their first contact with Twitter was at a conference.
Even if the features of Twitter are widely known, we will describe them in short:
Twitter allows registered users to share short posts with a maximum of 140 characters
to anybody else registered on Twitter in principle. Such posts are called “tweets” and

1
Corresponding Author (martin.ebner@tugraz.at)
they can be seen as public available SMS. On 22nd of February 2010 Twitter hit 50
millions tweets per day2. This overwhelming number of individual tweets evokes need
for organizing and selecting relevant tweets: Although Twitter was designed to be an
one-way information sharing channel, there were soon community-driven approaches
to enhance the service with mark-ups that allow for communication amongst Twitter
users [3].
Therefore, Twitter is equipped with a number of features:
(a) one can “follow”, that means select the streams of interesting Twitter user
(b) one can search for terms or tags (marked with “#”; hashtag) used within tweets
(c) one can directly address other users by a public reply (marked with “@” befor
the name of the other user; e.g. @mebner) or via a private direct message
(marked with “d”)
(d) one can cite and copy interesting tweets by “retweeting” them (marked with
“RT”; see [4] for more information on retweeting messages in Twitter).
The growing success of Twitter draws the attention of other systems on it. Big
social networking platforms like Facebook or Linkedin allow the integration of tweets
via a provided API and therefore make the application even more attractive. Twitter is
not just an additional communication channel; it seems it changed the way people are
exchanging information, links or dealing with social media in general. The limitation
to a message size of 140 characters is often argued to be both strength and weakness
of the microblogging service.
Twitter’s initiating question “What are you doing?” to motivate people to talk
about private stuff has already been expanded by “What is happening?” for
participating from business events, especially in the scope of media and e-learning.
The special Twitter syntax (see (a) to (d)) suggests being obviously useful for
scientific conferences. In order to join the community’s discussion about a topic or
the whole conference it is sufficient to tag one’s messages with the official hashtag of
the conference – a certain word following the # sign – or other existing tags. Without
knowing other people at the conference, it is therefore very easy to aggregate their
tweets. It becomes very simple to share ideas, impressions, comments and additional
materials about the conference on such a filtered #channel. Furthermore, there is no
setup required for the service. People, which are interested in actively using Twitter,
need to be registered with the service and post their messages via the Twitter website
or via one of the numerous clients using the provided API. Not only participants, but
also non-attendees can search for tweets within the conference hashtag and are
therefore able to follow the ongoing microblogging stream from the conference.
First, Twitter usage at conference was more or less something for early adopters,
playing with this new tool. Twitter usage was not officially announced or supported
by conference organization and therefore the communication at Twitter can be
described as a more or less hidden backchannel [5]. “Hidden” in this way does not
mean, that it was not available for everyone, but that it was unknown by most of other
participants. Especially at tech and media conferences the public announcement of a
hashtag and a short explanation of twitter became quite normal [6].
Besides this unofficial usage of Twitter at conference, there are several intended
scenarios conceivable that had already been used:

2
http://mashable.com/2010/02/22/twitter-50-million-tweets/ (last visited March 2010)
• Make the Twitter communication visible to all participants: To do so a public
extra screen with a Twitter wall, listening all current tweets was placed at the
front of the conference hall beside the normal whiteboard. For example, a twitter
wall was used at the EdMedia conference 2008 [5]. In [7] we pointed out how
people are using Twitter at conferences, showing that different stakeholders use
Twitter before, during, and after a conference with differing purposes in mind.
The analysis of the usage of the wall shows a surprisingly high participation of
the conference attendees at the wall. Nevertheless, to the best to our knowledge,
no content analysis about the communication and information flow at the wall
was done.
• As for the presumption, that a Twitter wall without concrete ideas about its
functionality tends to distract the public and the presenters; the usage of the
Twitter wall seemed to be too low-key. Therefore, conference organizers tried to
build on the pros of the tool and try to focus the communication. For example, it
was asked to use Twitter as an additional channel to ask questions in panel
discussions (see Campus Innovation 2009 in Hamburg). Additionally, sometimes
presenters directly ask for feedback, open questions or comments via Twitter [5].
This is used to make presentations more interactive, especially amongst big
audience where normally a microphone would be needed by an attendant to
interact with the presenter.
• Besides traditional scientific conferences, barcamps attract researchers too, at
least within the tech and media related disciplines. Barcamps are a kind of
“unconference” [8] that works without prepared scheduled presentations and
presenters for the benefit of more spontaneous, interactive communications and
knowledge exchange in ad hoc installed stand-up presentations, workshops and
discussion groups. Barcamps build up on Web 2.0 technologies, open formats
and tools. For example, they are typically using wikis for registrations and social
networking platforms for invitation to and promotion of the event. Even more
interesting, the usage of Twitter seems more or less a matter of course for the
participants: For example the announcement of workshops are typically
(additional) alerted via Twitter; the tweets are normally presented in the lobby of
the conference.
• Additionally, Twitter can be actively used by the conference organizators
before and after the event: to promote the call for papers and event itself, to alert
the full program, the registration possibilities, and afterwards the publication of
the proceedings [6].
Nevertheless, these intended and “officially” announced and supported usage of
Twitter at a conference are currently not widely used in every discipline and
community of researchers.

As Twitter is not limited to conference participants, it may evoke external


participants. Additionally, Twitter is sometimes actively used to report external
participants about the conference.

To sum up, there are few main several possibilities how Twitter is used in reality
according to conferences:
(a) for communication amongst participants
(b) for communication amongst organizers/presenters and audience
(c) for reporting to non-participants about the conference
Recent research work is mainly dealing with simple quantitative analyses, dealing
questions such as how many tweets from how many users in a specific time frame has
been contributed. This kind of benchmarking is used to address whether the usage of
Twitter was successful or not. In this publication we answer the main research
question: “Is Twitter proper to report from a conference in order to share the event
with the scientific community from outside the conference?” To do so we analyzed the
Twitter stream of Educamp 2010 in Hamburg, Germany.

2 Study

Thanks to of the Twitter API there are not only numerous clients for sending and
displaying tweets, but it also allows for the systematic analysis of the contents people
publish on the platform. In [9] we introduced an application that did a basic analysis
of tweets from communities of interest and visualized the main structural statistics
about those communities as well as a dynamic representation of the community’s
communication topics.
For this study we chose the EduCamp 2010 in Hamburg, due to the fact that many
participants are active Twitter users. Furthermore the organization committee
promoted the use of the microblogging platform by providing an appropriate hashtag
(#ec10hh) and encouraged to exchange and communicate via Twitter. As the
EduCamp is organized as a “barcamp” or “unconference” no scheduled program with
concrete talks or workshops was available before it starts.
We monitored the output of the Twitter usage between the 5th of February to the 4th of
March 2010. The EduCamp itself took place on 5th and 6th of February 2010. So in our
study also tweets from after the event are counted and analyzed. In summary 2.110
single tweets containing the mentioned hashtag are part of this study.
Focusing on our main research question, we took a closer look at the content of
each single tweet. Will a non-participant be able to follow the conference by watching
the live-stream or is he/she simply lost in information overflow? Is the information
provided on Twitter useful for the followers on the World Wide Web without being
there at the (un)conference?
In this study a two-stage analyses was chosen – first an automatic one by analyzing
main keywords, a so-called Formal Concept Analysis (FCA) (a detailed description
follows later on). Secondly tweets were analysed manually and categorised into the
following four categories:
• Irrelevant tweets
• Administrative tweets
• Relevant discussions
• Relevant tweets
In the following the categories are described more precisely:
Irrelevant Tweets

This category contains all tweets that are not relevant to the topic of the
conference. This means there is no hint towards learning, educational technology and
so on. These tweets are comparable to usual small talk on conferences and the Web.
Typical tweets in this category are the following:
• “Who found my drinking cup? #ec10hh”
• “was online #ec10hh”

Administrative Tweets

This category consists of all tweets with an administrative meaning. Example of


such content would be the room in which a session is going to take place, or any
technical information like wireless LAN or hotspot issues, video/audio streaming
issues etc. Typical tweets belonging to this category are the following:
• “Wifi is not available #ec10hh”
• “Session: new music interfaces at 5:00 pm in room Durkheim #ec10hh”
• “#ec10hh streaming in room humboldt not working?”

Relevant Discussions

This category counts all tweets that are relevant in the context of learning and
additionally are part of discussions on certain topics or replies to other participants.
Typical tweets of this category are tweets like:
• “@mccab99 tell me about the relationship between technology and
education? #ec10hh”
• “nice idea of @estudyskills Aggregation of all student weblogs at
Tumblelog - gives Overview. #ec10hh”
• “dito! rt@ lisarosa dance education: @mons7 shows, what we need:
creativity + passion + engagement. #ec10hh”

Relevant Tweets

This category consists of all tweets relevant in the context of e-learning and that
carries some valuable information with a strong focus on the conference topic. For
example, in this category tweets consist of a little description and an additional link to
another more detailed resource on the Web or simply describe / announce conference
outcomes. Typical tweets of this category are the following:
• “open-learning: initiative on OER usage for informal education:
http://u.nu/4a7y4 #ec10hh”
• “Session on exploring teaching and learning. Make a look to this article of
Gabi Reinmann: http://u.nu/3g8y4 #ec10hh”
• “portal of TU Braunschweig integrates numerous services like StudIP,
library, plus community: http://bit.ly/atxvaR #ec10hh”
The first two categories can be summed up in order to represent all the tweets of no
relevance in the context of e-learning or to the discussed conference topics, while the
latter two categories can be summed up in order to represent the relevant tweets that
might enhance the knowledge of external followers.

Results

Automated Analyses - Formal Concept Analysis (FCA)

Our first approach uses Formal Concept Analysis (FCA) in order to categorize twitter
users who write tweets about the same topics. A formal context is defined as set
structure К := (G, M, I), where G represents objects (in German “Gegenstände”), M
refers to attributes (in German “Merkmale”) and I describes a binary relation between
G and M [10] [11]. It can be represented as a matrix where rows containing objects
and columns containing attributes. If there is a binary relation between a certain row
and a certain column, this would be expressed by the identified cell having the value
of 1. If there is no relation the identified cell has the value 0.
A formal concept of a formal context К := (G, M, I) is defined as a pair (A, B) where
A is a subset of G, B is a subset of M (A = B’ and B = A’). A is called the extent and
B is called the intent of the formal concept (A, B). [10] The extent holds all objects
belonging to a certain concept. The intent contains all attributes (e.g. properties,
meanings) which apply to all those objects. As an entry point for further details on
Formal Concept Analysis we refer to [10] [11].
In this study we focus on categorizing different twitter users depending on the tweets
they wrote. The formal context К := (G, M, I) is defined as follows: All the keywords
used within the tweets represent the attributes (M) and the users who wrote these
tweets represent the objects (G). A concept (A, B) is represented by a set A of
keywords, which represents the intent and a set of twitter users (B) which represents
the extent of this concept (e.g. all the twitter users who used the keyword “e-learning”
in their tweets).
In order to extract the keywords of the tweets we used the Yahoo Term Extraction
Web Service3. The web service delivers a list of keywords for given tweets. For every
twitter user tweeting in the #ec10hh twitter stream the keywords were extracted and
as a result a formal context was created. Analysing the formal context was done with
ConExp4. ConExp allows users to explore, analyse and visualize formal concepts of a
formal context.
In order to focus only on the conference context irrelevant keywords were
manually deactivated from relevant ones. The result was a sparse matrix that means
that the extracted keywords did not overlap in a high manner. It can be interpreted in
such a way that monitored twitter users wrote about a lot of different topics (e.g. using
a lot of different keywords in their tweets). Due to the fact the categorization of
3
http://developer.yahoo.com/search/content/V1/termExtraction.html (last visited March 2010)
4
http://conexp.sourceforge.net (last visited March 2010)
twitter users based on their tweets using FCA did not result in valuable categories. It
can be stated that the examined tweets have high diversity and nearly no overlapping
keywords. This results that the tweets have to be analysed manually.

Manual analysis

In total 57% of 2.110 analysed tweets could generally categorized as irrelevant in


the context of the (un)conference outcome. As illustrated in Table 1 13% of the tweets
were categorized as administrative tweets, 9% were marked as discussions and 20%
were categorized as relevant tweets.

Total
Relevant Relevant Administrative Irrelevant Number
Tweets Discussions Tweets Tweets of
Tweets
Number of
Tweets including 456 201 278 1.175 2.110
Retweets
21,61% 9,53% 13,18% 55,69% 100,00%
Number of
Tweets not
containing "RT 346 157 227 996 1.726
@" or "via @" -
Clean Tweets
20,05% 9,10% 13,15% 57,71% 100,00%
Table 1 Categorization of tweets

In the following we are going to go into a more detailed analysis of the categories
relevant tweets and relevant discussions.

Number of
Number of Number of Tweets
Number of
Tweets Tweets not
Number Number of Tweets
Number of Number of containing not containing
of Tweets Number containing
Tweets Tweets "RT @" or containing "RT @" or
Category Tweets containing of "RT @"
containing containing containing "RT @" or "via @"
including at least Replies and
"RT @" "via @" "via @" "via @" - and
Retweets one link containing
but not Clean containing
"via @"
both Tweets at least
one link

Relevant 456 249 40 101 10 1 110 346 175


Tweets (349) (164) (32) (72) (7) (1) (78) (271) (120)
Relevant 201 37 70 42 2 0 44 157 28
Discussions (162) (19) (61) (33) (2) (0) (35) (127) (15)
Administrative 278 123 34 50 1 0 51 227 84
Tweets (193) (85) (26) (34) (1) (0) (35s) (158) (57)
Irrelevant 1.175 207 228 172 8 1 179 996 148
Tweets (939) (138) (179) (131) (8) (1) (138) (801) (102)
Total Number 2.110 616 372 365 21 2 384 1.726 435
of Tweets (1.643) (406) (298) (270) (18) (2) (286) (1.357) (294)

Table 2 Detailed analyses of Tweets


Table 2 displays all categorized tweets over the monitored time period.
Furthermore the table points also out all posts done during the conference time
(February 5th to February 6th 2010) in brackets.

Discussion

The analyses of the categories of interest the following crucial facts are to be
pointed out:
• In the timeslot of the conference (5th to 6th February 2010) 1.643 tweets
from 2.110 were tweeted by 272 users overall. Therefore it can be stated
that on average each user posted 6 tweets during the conference period.
During the analyses it was also attempted to separate participating active
Twitter user from non-participating ones, but because many user using
their accounts on an anonymous basis more than 100 users could not be
identified.
• Monitoring the conference: For a non-participant of a conference who
likes to monitor the event by checking the Twitter live stream it turned out
to be a hard job. Based on the assumption that even relevant tweets
containing simple statements are senseless without any distinct context
they are taking place only messages containing additional material (like
pictures, videos, or similar) are of interest. With other words if a link to
such attachment occurs it might be possible to understand the current
conference situation. Only 175 out of 2.110 tweets can offer such a
possibility. 8% or only one out of twelve is maybe of interest. This makes
following the conference stream a hard job. If we reduce the tweets to
tweets occurring during the conference period (5th and 6th February)
(Table 2) only 120 posts are of interest at all, which means about 6%.
• Retweets: Table 2 shows the high number of retweets (RT) in conference
twittering. In summary 384 RT occurred which is an overall of 18%. But
if the “relevant” category is analyzed towards RTs we see that 24% of all
tweets are simply a copy of another one. On one hand the retweet seems
to be a relevant message otherwise it would not be multiplied by another
user, but on the other hand it is senseless for people who aren’t at this
conference because they need the context of the proposed tweet to
understand the content.
Bearing in mind that RTs are a good instrument for pointing out the
importance of a tweet or helping to reach more Twitter user in order to
spread the world with announcements, they are not helpful if anyone is
interested on the content/output of the conference.
• The FCA showed after the keyword extraction and calculation that there
is no direct correlation amongst the conference participants. With other
words with this method it was not possible to get an overview about the
conference topics, the main assumptions or statements.
Conclusion

Our analysis showed, that the idea of Twitter usage for distributing or explaining
(un)conference topics, discussions or results to a broader public seems to be limited.
To interpret the data correctly, further comparative data and analysis are needed:
Our approach of analysing tweets at a conference was just a first step into a complex
research field.
As our analysis showed that the Twitter stream has a limited usefulness for
external participants that wanted to follow the conference from outside, our own ideas
and implicit theories about the Twitter usage should be modified.
Possible interpretations of the results could be that the Twitter usage follows other
logics, e.g.
- usage as a backchannel for conference participants or even a subgroup at the
conference, that means to comment silently with limited understanding
possibilities for outsiders;
- usage of self promotion and profiling, that means that for example citations
of people at the conference with a possible high retweet rate are posted to
generate attention for the own profile
- document and illustrate connections, for example friendship,
acquaintanceship, contacts to others (“I know … and likes her/his idea”)
- usage as a public notepad to collect relevant ideas, quotes or links
- usage as evaluation tool, for example to collect quotes about the conference
and the satisfaction of participants for the organization committee
Further research activities should combine the quantitative, semi-automatic analysis
as we have done within this contribution with additional questioning and involvement
about the aims and goals of the Twitter user at a conference.
Nevertheless, Twitter is a “new” tool and therefore adaptations of new forms of
communications, e.g. etiquette for documentation of conference for outsiders or a
more focussed usage of tweets by the presenters or conference organisers could many
influence future usage.
Practically, the obviously limited usefulness of the tweets of the EduCamp
participants for interested external participants provokes new ideas for practical and
effective alternative usage of the tweets as well as the need of appropriated tools. For
example, the own micro postings could be seen as personal notes, as they are
obviously easier interpreted and meaningful for the writers themselves.

References

1. Ebner, M., Lienhardt, C., Rohs M., and Meyer, I. (2010), Microblogs in higher education –
a chance to facilitate informal and process oriented learning? Computers & Education
2. Ebner, M; Schiefner, M.; (2008): Microblogging - more than fun?; in Proceedings of IADIS
Mobile Learning Conference 2008, Inmaculada Arnedillo Sánchez and Pedro Isaías ed.,
Portugal, 2008, p. 155-159
3. Java, A., Song, X., Finin, T. and B. Tseng (2007), Why we twitter: understanding mi-
croblogging usage and communities. In Proceedings of the 9th WebKDD and 1st SNA-
KDD 2007 workshop on Web mining and social network analysis, pages 56– 65. ACM,
2007.
4. Boyd, D., Golder, S. and Lotan, G (2010) Tweet, tweet, retweet: Conversational aspects of
retweeting on twitter. In Proceedings of the HICSS-43 Conference, Jan- uary 2010.
5. Ebner, M., (2009) Introducing Live Microblogging: How Single Presentations Can Be
Enhanced by the Mass, Journal of Research in Innovative Teaching (JRIT), 2 (1), p. 91-
100, 2009
6. Ebner, M., Reinhardt, W.: Social networking in scientific conferences – Twitter as tool for
strengthen a scientific community. In: Proceedings of the 1st International Workshop on
Science 2.0 for TEL. 2009.
7. Reinhardt, W., Ebner, M., Beham, G., Costa, C.: How people are using Twitter during
Conferences. In: Hornung-Prähauser, V., Luckmann, M. (Eds.). Creativity and Innovation
Competencies on the Web. Proceedings of the 5th EduMedia 2009, Salzburg, pages 145-
156. 2009.
8. Bernhardt, Thomas / Kirchner, Marcel (2009): Web 2.0 meets conference – the EduCamp
as a new format of participation and exchange in the world of education. In: Ebner, Martin /
Schiefner, Mandy (Hrsg.): Looking Toward the Future of Technology-Enhanced
Education: Ubiquitous Learning and the Digital Native. IGI Global, Hershey.
9. Reinhardt, W.: Visualizing the dynamics of communication of Communities of Practice on
Twitter. In: Proceedings of the 3rd International Workshop on Building Technology
Enhanced Learning solutions for Communities of Practice, 2009.
10. Wille, R. (2005): Formal Concept Analysis as Mathematical Theory of Concepts and
Concept Hierarchies. In: Formal Concept Analysis. Springer LNAI 3626, pages: 1-33.
11. Ganter, B. & Wille, R. (1999), Formal concept analysis: Mathematical foundations,
Springer, Berlin-Heidelberg.
12. Costa, C., Beham, G., Reinhardt, W., Sillaots, M.: Microblogging in Technology Enhanced
Learning: A Use-Case Inspection of PPE Summer School 2008. In: Proceedings of the 2nd
SIRTEL workshop on Social Information Retrieval for Technology Enhanced Learning.
2008.
13. Grosseck, G. , Holotescu, C. (2008), Can we use twitter for educational activities? In
Proceedings of the 4th International Scientific Conference eLSE ”eLearning and Software
for Education”, April 2008

You might also like