Content uploaded by Martin Ebner
Author content
All content in this area was uploaded by Martin Ebner on Jun 03, 2016
Content may be subject to copyright.
TwitterSuitcase – How to Make Twitter Useful for Event/Lecture Participants
Draft – originally published in: Ebner. M., Harmandic, S. (2016) TwitterSuitcase – How to Make Twitter Useful
for Event/Lecture Participants. In: Wallace, K. (Ed.). Learning Environments: Emerging Theories, Applications
and Future Directions, Nova publishers, pp. 175 – 196, ISBN 978-1-63484-893-0
TwitterSuitcase – How to Make Twitter Useful
for Event/Lecture Participants
Martin Ebner1
Sead Harmandic2
Educational Technology
Graz University of Technology
Austria
10/01/2016
Keywords: Twitter, Microblogging, Events, analysis, visualisation, education
1 martin.ebner@tugraz.at (corresponding author)
2 sead.harmandic@tugraz.at
Martin Ebner & Sead Harmandic
Abstract. This publication introduces a tool for analysis and visualization of tweets for a
particular event. Recently, many related studies have analysed and extracted information
from social media such as Twitter. However, the majority of those approaches are aiming at
providing specific content-based information. The objective of this research work is to
provide an overview of data collection for a specific event by analysing and visualizing all
collected tweets. Furthermore, this paper will focus on applying the outlined approach when
using Twitter during conferences and determining what kind of information is used for some
specific event.
Two Twitter events have been created for this purpose and the obtained results are presented,
explained and discussed. The result of data processing represents a summary of a Twitter
event and allows an overview of various information such as the most popular hashtags, users
who tweet the most, the most published links, list of all used software platforms, etc. It also
includes a timeline of tweets in terms of years, months, days and hours depending on duration
of events. Finally, the future of Twitter and the visualization of its data is discussed.
1 Introduction+
In year 2005, Tim O'Reilly (O’Reilly, 2005) described the foundations of Web 2.0 and the
meaning of the term itself. He established a new user's attitude towards the web and web
content by exploiting the possibilities of blogging services and without merely relying on
information sources controlled by a small group of people (such as editors). This statement
was also backed up by Dan Gillmor in his book "`We the Media"', where he discussed the
phenomenon of internet journalists (bloggers) and the way they handle and publish
information (Gillmor, 2004). In this period, blogging became very popular not only because
of its simplicity, but because it established the term Blogging as one of the mainstream
information sources. So, what is Blogging and where does it come from? According to
Walker (2005) a weblog, or just blog, is described as a frequently updated website consisting
of dated entries arranged in a reverse chronological order so that the most recent post appears
first. Hence, blogging describes activities related to adding new entries to a blog or
maintaining the existing ones. The additional popularity that the use of weblogs gained is
related to indirect communication with a large amount of people who are often called social
network or social community (Schiefner & Ebner, 2008). It allows anyone to become an
(in)active member of a social network without of preconditions. A user's contribution to a
network determines whether he will be an active or inactive member.
In 2006, a new online social network platform called Twitter was launched, thus further
increasing the popularity of microblogging. The number of active users reached 140 million
in the first six years, until 2012, with an upward tendency (Twitter, 2015) The upward trend
was confirmed in 2015 when the total number of active users reached over 300 million,
according to Welch (Welch, 2015). The main objective of this research work is to analyse
Twitter data during specific events such as conferences or lectures and to visualize them in a
form that can be understood by a common user. The results of the survey will be discussed
and possible improvements to the data processing will be proposed.
1.1 Microblogging+
Microblogging is a new form of blogging where users can publish various information, taking
into account specific content restrictions. Kaplan and Haenlein (2011) described
microblogging as a text with a maximum of 200 characters, distributed via instant messages,
email, cell phones or the web, containing mainly short sentences, HTTP links or images.
Regular weblogs are mainly used for sharing personal thoughts, saving knowledge and
discussion. On the other hand, microblogging is mostly used for a fast exchange of thoughts,
TwitterSuitcase – How to Make Twitter Useful for Event/Lecture Participants
ideas and information sharing (Beham et al, 2008). Although the amount of information is
clearly greater in case of weblogs, the simplicity and flexibility of microblogging makes it
more popular and in some cases even more user-friendly then weblogs.
1.2 Twitter+
Twitter is probably the most popular microblogging platform currently available. According
to Twitter3, the service has approximately 316 million active users sending over 500 million
tweets per day. One of the first studies on Twitter have been published under the title “Why
we Twitter: Understanding Microblogging Usage and Communities” with the special aim on
three different objectives (Java, 2007). The first objective was to analyse topological and
geographical properties of Twitter's network and the second objective was to analyse
intentions of a user in combination with community level. The third objective was how users
with same intentions interact with each other. Java (2007) stated that Twitter users,
depending on their intentions, could be categorized into four different types:
• Daily chatter: The common and the largest user group on Twitter. They are
using microblogging as a tracker of their daily routine.
• Conversations: Second largest user group with aproximately 21% of all
Twitter users who mostly reply to posts of other users.
• Sharing information/URLs: Sharing URLs of relevant resources including
short comments or hashtags. This group makes up about 13% of all posts.
• Reporting news: This group is created out of users who are frequently posting
information or comments about recent events. It is very popular among
different automatic services like weather forecast due to the accessibility of
the Twitter APIs.
The categorization of all users was divided into 3 main groups according to their
microblogging activities:
• Information Source: Group of users known for their valuable updates.
Although the frequency of updates may diversify between frequently or
infrequently, the quality of the content they are publishing makes them very
popular.
• Information Seeker: This group of users is posting rarely. Their main focus is
on reading and seeking for information posted by other users.
• Friends: A widespread group of users exploiting microblogging platforms as
social networks sites for extending relationships as well as for discussion
among followers. Typically, this group has also sub-categories such as family,
closest friends, co-workers etc. The categorization of friend on microblogging
platforms used as social network sites was also proposed by Andrew Lavallee
(2007) in his article called “Friends swap twitters, and frustration”.
2 State+of+the+Art+
Shortly after Twitter was launched in 2006 many researchers have started exploiting its
possibilities for various applications within the microblogging (Ebner, 2013).
3 https://about.twitter.com/company (last visited August 2015)
Martin Ebner & Sead Harmandic
In year 2008 Huberman et al. published their research called “Social Networks that Matter:
Twitter under the Microscope” where they have distinguish between the followers and the
hidden network of connections which underlies the “declared” set of friends (Huberman,
2008). According to Huberman et al. the term “friend” is defined as a person whom the user
has directed at least two posts. Using this definition, they were able to find out how many
friends each user really has and then compared this number to the number of followers
(subscriber of a user) and followees (people followed by a user). Their expectation that users
who have a large audiences or many followers will post more often than users whose
audience is rather small was confirmed as can be seen in Figure 1. Although the total amount
of tweets is increasing with the amount of followers and friends, it does not correspond to the
function of number of friends. This leads to an assumption that the number of friends should
be considered in order to predict the activity of a Twitter user, not the number of followers
(see Figure 2). Therefore, they have distinguished between users who actually communicate
using direct messages with each other and users who have a large number of followers but
less communication with them in order to determine which Twitter users are providing more
or less information. The results have shown that the majority of Twitter users interact with a
small group of friends. According to Huberman et al. 98.8\% of users have less friends than
followees.
Fig. 1 Number of posts as a function of the number of followers (Hubmann, 2008)
Fig. 2 Number of posts as a function of the number of followers (Hubmann, 2008)
In the paper "`Measuring User Influence in Twitter: The Million Follower Fallacy"' the
researcher has tried to estimate user dynamic influence across time and topics. Cha et al.
(2010) used a large amount of data collected from Twitter and categorized them into three
main groups for measuring the influence, such as indegree (determines the popularity of a
user), retweet (indicates the ability of a user to create valuable content) and mentions
(describes the ability of a user to engage or involve other users in conversation). The results
of this analysis have been created by using Spearman’s rank correlation coefficient to test the
strength of a relationship between two data sets. The test showed that the users are separated
into the following groups according to the calculated coefficient:
• The most followed users: These users have a high indegree coefficient and
belong to a wide variety of public figures and news sources. Such users are
politicians (Barack Obama), athletes (Shaquille O'Neal), news sources (New
York Times) and celebrities like actors, musicians and models (Asthon
Kutscher)
• The most retweeted users: This group of users are tracking current topics and
knowledgeable people in different fields. Such users are news sites (New York
Times) aggregation services (TwitterTipps) and businessmen (Bill Gates)
• The most mentioned users: Mostly celebrities who have been mentioned by
ordinary users.
The overlap of those three measures is shown as a normalized chart to the total of 100% in
Figure 3. The final conclusion of this research stated that the popularity of a user does not
necessarily lead to a greater influence factor of the same user. Though the number of
followers does provide a better position, the user must exert a greater personal effort to gain
influence.
TwitterSuitcase – How to Make Twitter Useful for Event/Lecture Participants
Fig. 3 Venn diagram of the top 100 influentials across measures (Cha et al, 2010)
The use of Twitter for establishing correlation between the tweet content and its geotag is
described by Hiruta et al. (2010) in their research called “Detection, Classification and
Visualization of Place-triggered Geotagged Tweets”. They have proposed and evaluated a
method to recognize and classify tweets triggered by specific places where the user is located.
The research is based on an assumption that the real world is structured as a collection of
descriptive attributes. In order to exploit such assumption a system that can extract, classify
and provide real-time dynamic attributes for specific searched event must be provided.
Twitter has been used as data source for this research mostly because of its public and agile
nature as a communication medium. The first step in this research was to detect place-
triggered geotagged tweets and determine whether the tweet contains location-based
information. In the second step, those tweets were classified in accordance with keywords
and regular expressions. The process of classification itself is divided into five different
types, such as:
• Report of whereabouts: Tweets referring to user's current location.
• Food: Tweet containing information about food or drink.
• Weather: Tweet about the weather at the current location.
• Back at home: Tweet containing straightforward information that the entity
(mostly a person) came back home.
• Earthquake: Tweet referring to a recent earthquake in its area
These classification types are detected by filter modules. The first hypothesis is that the tweet
is of whereabouts type and most of those tweets come from various check-in services (such
as Foursquare) so the location could be extracted precisely. Other four filters (food, weather,
back at home and earthquake) are using a keyword matching approach combined with online
dictionaries. If the text in a tweet contains the words found in the synonym list of four filters
(for example food), the tweet is considered to be of a type considered appropriate (in this case
tweet is of type food). A single tweet, based on its content, can be classified with more than
one type. The approach was tested by 18 people, who have been asked to manually classify
approximately 4,2 millions of tweets gathered between January and March 2012. Although
this research and its implementation are experimental, the results are quite encouraging due to
the achieved accuracy of 82% by 18 human classifiers.
Analyzing and detecting various content objectives within tweets is showing great potential
and there are many examples of applications. Such example is described by Sakaki et al.
(2010) in their research “Earthquake Shakes Twitter Users: Real-time Event Detection by
Social Sensors”. In this paper they have designed a system, which can classify specific tweets
in order to determine recent earthquakes around the world. Since Japan has a long history of
numerous earthquakes and the popularity of Twitter is very high, they have established a
functioning system to detect tweets concerning earthquakes and issue a warning to the
impacted areas. The second reason for using Japan as a test region, other than a large number
of active users, is a good distribution of users across the country. The basic assumption of the
researcher was that every Twitter user will be considered a possible earthquake sensor and
each tweet from this sensor will be regarded as sensor information also known as social
sensor. The classification of tweets gave the possibility to distinguish between relevant and
irrelevant tweets. Hence, relevant tweets are only to be considered using the aforementioned
Martin Ebner & Sead Harmandic
basic assumption and an assumption that each tweet is associated with its latitude and
longitude describing time and location. The experimental model of this research proved to be
very successful during a test phase in 2009, so in August 2010 the Earthquake Reporting
System was fully operational. An earthquake early warning service was already launched by
the government in 2007, which gave the possibility to detect primary waves of seismic
activities and to calculate possible earthquake arrivals. A combination of these two services
gives the possibility of using earthquake warnings on a completely new level of news
broadcasting. However, the visualization of such service needs to be improved with respect to
the appropriate area maps and locations including all earthquake information.
In their research, Pak and Paroubek (2010) set focus on sentiment analysis in tweets. The
objective of their research was to automatically collect data for sentiment analysis and
opinion mining. The classification of the tweets created three main categories, such as
positive (containing happy emoticons), negative (containing sad emoticons) and neutral (no
emoticons at all). The sentiment (neutral) was used as a training set for sentiment classifier,
since they have determined that neutral posts, such as posts from New York Times or
Washington Post, are a good choice to start training classifiers. The foundation of sentiment
classifier was built on multinomial Naive-Bayes Classifier using N-Grams and Part-Of-
Speech Tags. Their research was based on a previously covered topic by Jansen et al. (2009)
on microblogging becoming a more popular way of communication and its identification as
an electronic word-of-mouth when it comes to sharing customer opinions towards various
brands. In the research of Jansen et al. (2009) they have discovered that evaluation of
microblogging posts could influence the rating of various brands. Apparently, 19% of all
microblogs mention a brand and about 20% of those posts contain some kind of an
expression. Over 50% of those expressions show positive and about 33% show negative
attitudes towards a brand. Although they have created useful information, the visual
representation of those results could be improved.
The use of microblogging services was discussed in the work of Beham et al. (2008) in their
study “How People are using Twitter during Conferences”. They have analysed how the use
of a special hashtag for Twitter before, during and after a conference can be exploited, what
the motives for tweeting during a conference are and finally what value that information
carries. The use of Twitter was divided into three different stages of a conference such as
Before, During and After a conference. The participants of the study had to complete a
survey consisting of 34 questions. The subjects were required to answer if they already had a
Twitter account, if they use it for professional or private purposes or both, if they use Twitter
to actively communicate during conferences, etc. The results have shown that 95.1% of users
already had a Twitter account and they have been using it both for personal and private
reasons. Interestingly, about 51% of users were using the same “single account for multiple
use cases” approach with other communication tools as well. Over two-thirds or 67.5% of
Twitter users tweeted actively during a conference. The content of tweeted text was also
analysed and the majority of the content or 50% was defined as a plain text (without any links
or images). Approximately 10% of tweets contained links to external services and were
mainly sent by the conference delegates. The survey also included questions about the
expectations and attitudes towards using Twitter during a conference, which was rather
skeptically at the beginning but has changed in the end especially throughout discussion,
spreading and sharing conference related information.
The final results of this research have shown that discussion on various topics is not limited
only to the face-to-face audience and could be exploited effectively. There were also some
TwitterSuitcase – How to Make Twitter Useful for Event/Lecture Participants
disadvantages like impracticality to work with data formats other than plain text and web
links.
3 Motivation+
Twitter is a real-time communication social network used for news updates, information
sharing and discussion on various topics. Therefore, a large amount of data is circulating
within the social network and it requires a deeper analysis in order to gain more useful
information. The lack of possibility to save tweets for further analysis is the biggest problem
when working with Twitter, since the data stream cannot be stopped nor can the old tweets be
viewed. Once a particular event is over, it is very difficult to find, analyse and visualize all
the tweets corresponding to the event. During a conference or lecture, most of the participants
have their focus directed at a presentation or discussion being held while ignoring Twitter and
its content. Beham et al. (2008) have stated out that approximately 95% of all participants
already have a Twitter account. This implied that a majority of participants will be using
Twitter once a presentation is over, in order to catch up on the missed updates by their
friends, followers or followees. Basically, the bigger the community, the more information
flows and this increases the probability of missing important conference-based or lecture-
based information. For a common participant, it is important to have access to all kind of data
at any time. This kind of information includes posted updates, HTTP links, images, the most
popular hashtags, and so on. Therefore, there is a necessity for a tool that can provide such
information event though the event/lecture is over. This leads us to the research question:
“What kinds of conference-based information are we capable to providing during and after
some tweet event?”. In this research work a tool is developed and presented, that fulfil these
needs (called TwitterSuitcase).
4 Prototype+Development+
Storing and analyzing tweets during some event was successfully implemented by Thomas
Altmann in his master’s thesis (Altmann, 2014). The process was divided into three services
such as TweetCollector, TwitterStat4 and TwitterWall5. TweetCollector builds up a central
application by collecting all tweets containing a specific hashtag (e.g. all Tweets containing
#news hashtag). TwitterStat is a statistical extension of previously collected tweets that
provides information about collected tweets such as mentions, retweets and so on.
TwitterWall is another extension of TweetCollector used for presenting and displaying the
collected tweets. It also provides a possibility to search for specific hashtags or words within
gathered data. Now a new tool called TwitterSuitcase is created to fulfil the needs for analysis
and visualization of Twitter event/lecture data collected by TweetCollector.
4.1 TwitterSuitcase+
TwitterSuitcase is a tool created for visualization of specific events. The functionality of
TwitterSuitcase is based on a previously gathered tweet collection, accomplished by
TweetCollector. The provided tweets have been analysed and parsed into the total of eight
visualization categories in accordance with visualization objectives. The categories include
breaking tweets into single words, extraction of all links, creation of screenshots, view of
tweet's timeline, extraction of all hashtags and other relevant information.
4 http://twitter.learninglab.tugraz.at/stat/ (last visited January 2016)
5 http://twitter.learninglab.tugraz.at/wall/ (last visited January 2016)
Martin Ebner & Sead Harmandic
Top Users is the first visualization category created from the list of all Twitter users who
have written about the given hashtag (see Figure 4a). Their activities are represented in a
form of a word cloud. This means that the more tweets each user has, the bigger the
representation of his name will be.
Top Links presents the list of the popular HTTP links, including the number of occurrences
within an event (see Figure 4b). The default preview is restricted to a maximum of twenty
links, but there is also a possibility to view all the links that appeared in relation to an event.
Most Popular Retweets corresponds to the list of tweet appearances within an event (see
Figure 5a). As in the case of Top Links, the default preview is restricted to a total of fifty
tweets including the option to see all the collected tweets.
Timeline of Tweets is a visual representation of collected tweets according to date and time
when they have been created or sent (see Figure 5b). Tweets are grouped by years, months,
days and hours and the two largest groups having more than two elements are displayed. The
groups are displayed according to the following sorting rule:
Year > Month > Day
Fig. 4a Overview of Top Users
Fig. 4b Overview of Top Links
Fig. 5a Overview of Most popular Retweets
Fig. 5b Timline of Tweets
Top Words display the most popular words apart from hashtags found and counted during a
specific event in the form of a pie chart (see Figure 6a).
Top Software presents top twenty software or applications used to send a tweet (see Figure
6b).
Fig. 6a Overview of Top Words
Fig. 6b Overview of Top Software
Most popular Hashtags counts the occurrences of all hashtags, which appear within an
event. Top thirty hashtags are presented in a chart shown in Figure 7. Although all hashtags
have been taken into account, it is not possible to display all of them because of the screen
limitations.
Fig. 7 Top Hashtags
Top Screenshots are snapshots of the most popular http links. The basis for snapshots is
provided by Google API and is called PageSpeed6. Fig. 8 shows the created snapshots
Fig. 8 Top Screenhots
Wikipedia links are created from the most popular hashtags using Wikipedia search API (see
Figure 9. The results are classified into partial and full results, depending on the sort of
information they are retrieving. Partial results provide a list of possible referrers or article
6 https://developers.google.com/speed/pagespeed/insights/ (last visited January 2016)
TwitterSuitcase – How to Make Twitter Useful for Event/Lecture Participants
recommendations, but do not have the main Wikipedia article. Full results automatically lead
to the main Wikipedia article handling the given hashtag as a subject.
Fig. 9 Wikipeadia Links
Show all Tweets and Show all Links in a new window display all the tweets and links
collected during a specific event.
There is also a possibility of embedding the collected information into external web
applications or further data processing by exploiting the available API which works with
JSON data format. Listing 1 shows what a response for an API call can look like.
[{"id":"79",
"keyword":"#iaaf",
"title":"IAAF Event Title ",
"count":"906"},
{"id":"80",
"keyword":"#beijing2015",
"title":"Athletic World Championship Beijing 2015",
"count":"2869"},
{"id":"81",
"keyword":"#football",
"title":"Football TwitterEvent 2015",
"count":"8991"}]
Listing 1 JSON Response to API Call
5 Evaluation+
The next step, which will be realized in this section, is to evaluate the real-time tweets for a
specific event in order to verify the functionality of the presented tool. The following
hashtags #edmediaconf and #emoocs2014 have been chosen for building up the test
bundles.
5.1 #edmediaconf+
Why the hashatg #edmediaconf? It is one of the most popular international conferences on
educational media and technology organized by AACE7. Since the conference is dealing with
various topics on technology used in education, it is quite understandable that Twitter will be
one of the most, when not the most used microblogging services for information publishing
and sharing.
As was mentioned in section 4, the basis for gathering tweets for a specific event was
provided by TweetCollector. The total number of collected tweets was 443 for the period
between June 16th and September 29th, 2014.
The word cloud of users who write the most displays all users who have sent a tweet
containing #edmediaconf hashtag. The names and number of tweets of users who have
twenty or more tweets and are placed in the top five category are: awengblom(39), aace(28),
mebner(27), amysampsonuk(23) and diando70(22). These top five users have published 139
7 http://www.aace.org/conf/edmedia/ (last visited January 2016)
Martin Ebner & Sead Harmandic
tweets, which makes about 32.1% of collected tweets and the top ten users are providing 211
tweets (about 48.8% of the total number of tweets). Top five of the most popular links are:
• http://t.co/I3au4Aweiy (18)
• http://tec-variety.com (5)
• http://t.co/hnlJGIuCtg (5)
• http://t.co/YBUIg0U3sR (5)
• http://t.co/40P9JPLVkF (4)
The popularity of links correlates corresponds to the list of the most popular retweets (see
Figure 10a), which shows the tweets with the most retweets by other users. The most popular
tweet by far had 17 retweets and was published by user “mebner” (see Figure 11).
Fig. 10a The Most Popular Retweets for #edmediaconf
Fig. 10b Snapshot of a tweet by “mebner” for #edmediaconf
Fig. 11 The most popular tweet for #edmediaconf by “mebner”
The tweets were collected between 17th of June and 29th of September 2014. While the
majority of tweets were collected in June (419 or 96.8%), a small amount of tweets were
published in July, August and September (14 or 3.2%). Figure 12 shows what a graph of
daily collected tweets looks like. The most popular words are actually prepositions (such as
about, in, to, at, etc.), articles (such as the, a, etc.) or Twitter's “RT” shortcut for retweet.
However, there are also some other information such as: learning (82 occurrences), aace (69
occurrences) and tampere (36 occurrences). According to the chart of top fifteen software or
applications, most of the tweets were sent from Apple devices (195 or 46.4%). Various web
applications or clients for Twitter have managed to take the second position with 144 tweets
or 34.2%. Android-based applications have been used to send 33 or 7.9% of tweets. The
remaining 48 tweets was divided between various software or applications which sent mostly
just a few tweets each, such as Foursquare, SharedBy, Twitter for Windows Phone, Instagram
and many more.
Fig. 12 Timeline of Tweets for #edmediaconf
The top ten hashtags from the most popular hashtags graph, besides \#edmediaconf with 433
occurrences, are:
• #oasisutafi (34)
• #learninganalytics (31)
• #edmedia (20)
• #edmedia2014 (18)
• #tampere (12)
• #edmediakeynote (6)
• #funproject (6)
• #aktiivi (5)
• #lecturecapture (4)
• #edtech (4)
The aforementioned hashtags are all related to the topics of the conference, such as
#oasisutafi (a social learning space and living lab at the University of Tampere8), #tampere
8 http://www.uta.fi/en/ (last visited January 2016)
TwitterSuitcase – How to Make Twitter Useful for Event/Lecture Participants
(location9 of EdMedia 2014 conference), #edmedia and #edmedia2014 (EdMedia conference
hashtags10), etc.
Top Screenshots are created out of the most popular links and offer an overview of tweets
using such links. Figure 10b shows an overview of the most popular tweet published by
“mebner”. The Wikipedia search API for the above mentioned hashtags returned only one
full article, for hashtag #tempere.
5.2 #emoocs2014+
Why the hastag #emoocs2014? It is a hashtag used for one of the greatest conferences
discussing massive open online courses. Therefore, the analysis and visualization of data
related to the event would be appropriate. The collection of tweets included 4450 tweets,
which were published between 10th of February and 6th of August 2014.
All the gathered tweets were grouped by users and are presented in a word cloud showed in
Figure 13a. Top ten users from the word cloud have published a total of 1138 or 25.6% of all
tweets, and their names along with the number of tweets are as follows: moocf(185),
Agora_Sup(141), fuscia_info(134), pabloachard(124), mooc24(120), tkoscielniak(103),
bobreuter(85), OpenEduEU(84), yveszieba(81) and redasadki(81).
Fig. 13a Top Users #emoocs2014
Fig. 13b Top Links #emoocs2014
Fig. 14 Most Popular Retwets #emoocs2014
The most popular links are shown in Figure 13b but this list does not necessarily correlate to
the list of the most popular retweets which is shown in Figure 14. Top five of the most
popular links are:
• http://bit.ly/1la3yJX (32)
• http://twubs.com/eMOOCs2014 (28)
• http://www.emoocs2014.eu/sites/default/files/Proceedings-Moocs-Summit-
2014.pdf (19)
• http://goo.gl/Ujaztf (16)
• http://www.openuped.eu (15)
When observing the list of the most popular retweets, we can determine that there is no
significant difference among the top ten results, since the difference between the first and
tenth tweet is only six occurrences, which is rather insignificant when considering the total
amount of tweets (being 4450).
While the majority of tweets was collected at the very beginning of collection period, which
was February 2014 with 4333 tweets or 97.3%, the remaining months from March until
August 2014 have given only 117 tweets or 2.7% of all tweets.
Figure 15a shows the twenty most popular words within a event. The best result was achieved
by a Twitter “RT” shortcut with 2567 occurrences. Other words worth mentioning are
moocs(802), mooc(639), learning(339) and openedueu(319). The rest of the words belong
mostly to prepositions or articles.
9 http://www.tampere.fi/english/ (last visited January 2016)
10 https://twitter.com/hashtag/edmedia (last visited January 2016)
Martin Ebner & Sead Harmandic
Top used software or application (see Figure 15b) shows that the majority of tweets, precisely
1574 or 35.4%, were sent directly from Twitter using web service. The second largest group
was represented through Apple devices (divided into various subgroups such as Twitter for
iPhone, iPad or Mac) with a total of 1253 tweets or 28.5%. The third position is taken by a
Twitter application called TweetDeck with 564 tweets or 12.7%, followed by Twitter for
Android in the fourth place with 288 tweets or 6.5\% of all tweets. The rest was split on
various applications such as HootSuite, Mobile Web, Tweet Button, etc.
Fig. 15a Top Words #emoocs2014
Fig. 15b Top Software #emoocs2014
The results for the hashtag popularity are listed below with the number of occurrences:
• #emoocs2014 (4435)
• #mooc (308)
• #moocs (270)
• #elearning (72)
• #futurelearn (60)
• #vtecl (48)
• #elearningpapers (48)
• #bigdata (45)
• #heie (42)
• #edchat (42)
The listing displays the top ten hashtags used within the event and are undoubtedly related to
the topics of conference. The last two sections include the visualization of the most popular
links into snapshots and there are three full articles for #elearning, #mooc and #futurelearn
hashtags on Wikipedia.
6 Discussion+
The use of TwitterSuitcase as a visualization tool enables a more detailed insight into an
event-related collection of tweets. It provides answers to questions, such as “Who wrote the
most tweets?”, “What are the most popular words, hashtags or links?”, “What links have been
posted?” and so on, simply by observing various parts of TwitterSuitcase, such as Top Users,
Most Popular Hashtags and so on. According to the results, it could be stated that during
events (such as #emoocs20014 and #edmediaconf) there is always a small group of Twitter
users who post the majority of tweets creating about 30% to 40% of all tweets. Such users
are not only the strongest microblogging mechanism pushing forward, but also the reason
why the majority of other users are also using the same microblogging platform.
Though the visual transformation of a group of tweets is made more comprehensible for an
ordinary user, it does not imply that there is no room for improvement. Although, the topic of
this paper is the visualization of tweets, it could be extended by performing a deeper data
analysis of tweets that would eventually lead to attempts to build up semantic triples in the
background and try to present them. A possible extension of TwitterSuitcase could be the
TwitterSuitcase – How to Make Twitter Useful for Event/Lecture Participants
location-based representation of tweets using maps for visual presentation such as Google
Maps. The use of mindmaps in order to show the relations between users and tweets might
also be useful.
7 Conclusion+
The future use of Twitter and other microblogging services is very encouraging since it's
gaining popularity each day. Asur and Huberman (2010) have stated the possibility of using
Twitter for predicting real world outcomes in the form of a box-office results for movies by
analysing almost three million tweets. Using microblogging services during conferences
revealed a great potential, but in order to exploit that potential to the maximum, there must be
a closer collaboration between developers and organizers of conferences. Extending services
is not the only condition to be fulfilled, because it all comes down to the participants and their
activity or inactivity on Twitter during the conference. This is how the amount of the
observed data can be enlarged and how we can provide better and more conference-related
information.
8 References+
Altmann, T. (2014). Potential of twitter archives. Master’s thesis, Graz University of
Technology.
Asur, S. and Huberman, B. A. (2010). Predicting the future with social media. International
Conference on Web Intelligence and Intelligent Agent Technology.
Beham, G., Ebner, M., Reinhardt, W. and Costa, C. (2008). How people are using twitter
during conferences. Proceeding of 5. EduMedia conference, pages 145–156.
Cha, M., Haddadi, H., Benevenuto, F., and Gummadi, K. P. (2010). Measuring user influence
in twitter: The million follower fallacy. Fourth International AAAI Conference on Weblogs
and Social Media.
Chris Welch, B. P. (2015). Twitter reaches 300 million active users, but the stock crashes
after earnings leak early. http://www.theverge.com/2015/4/ 28/8509855/twitter-earnings-q1-
2015-leak-selerity. Accessed: 2015-08-25.
Ebner, M. and Schiefner, M. (2008). Microblogging - more than fun? Procceding of IADIS
Mobile Learning Conference 2008, page 155–159.
Ebner, M. (2013). The Influence of Twitter on the Academic Environment. Patrut, B., Patrut,
M., Cmeciu, C. (ed.). Social Media and the New Academic Environment: Pedagogical
Challenges. IGI Global. 2013. pp. 293-307.
Gillmor, D. (2004). We the Media (Classique Us). O’Reilly and Asso- ciates.
Hiruta, S., Yonezawa, T., Jurmu, M., and Tokuda, H. (2010). De- tection, classification and
visualization of place-triggered geotagged tweets. Proceed- ings of the 2012 ACM
Conference on Ubiquitous Computing, page 956–963.
Huberman, B. A., Romero, D. M., and Wu, F. (2008). So- cial networks that matter: Twitter
under the microscope. Social Science Research Network.
Martin Ebner & Sead Harmandic
Jansen, B. J., Zhang, M., Sobel, K., and Chowdury, A. (2009). Twitter power: Tweets as
electronic word of mouth. Journal of the American society for information science and
technology, page 2169–2188.
Java, A., Song, X., Finin, T., and Tseng, B. (2007). Why we twitter: understanding
microblogging usage and communities. Proceedings of the 9th WebKDD and 1st SNA-KDD
2007 workshop on Web mining and social network analysis, pages 56–65.
Kaplan, A. M. and Haenlein, M. (2011). The early bird catches the news: Nine things you
should know about micro-blogging. Business Hori- zons, Vol. 54, pages 89–103.
Lavallee, A. (2007). Friends swap twitters, and frustration. http:
//www.wsj.com/articles/SB117373145818634482. Accessed: 2015-08-24.
O’Reilly, T. (2005). What is Web 2.0. O’Reilly Media, Inc.
Pak, A. and Paroubek, P. (2010). Twitter as a corpus for sentiment analysis and opinion
mining. In Proceedings of the Seventh Conference on International Language Resources and
Evaluation.
Rettberg, J. W. (2005). Weblog definition. http://jilltxt.net/?p= 227. Accessed: 2015-09-21.
Sakaki, T., Okazaki, M., and Matsuo, Y. (2010). Earthquake shakes twitter users: real-time
event detection by social sensors. Proceedings of the 19th international conference on World
wide web, pages 851–860.
Schiefner, M., Ebner, M. (2008) "Weblogs, more than just a toy?" or "Should I keep an e-
Portfolio for my PhD study?" In: Interactive Computer Aided Learning, ICL 2008, Villach
Twitter (2015). Twitter turns six. https://blog.twitter.com/2012/ twitter-turns-six. Accessed:
2015-08-25.
