=Paper=
{{Paper
|id=None
|storemode=property
|title=Citation Analysis in Twitter: Approaches for Defining and Measuring Information Flows within Tweets during Scientific Conferences
|pdfUrl=https://ceur-ws.org/Vol-718/paper_04.pdf
|volume=Vol-718
|dblpUrl=https://dblp.org/rec/conf/msm/WellerDP11
}}
==Citation Analysis in Twitter: Approaches for Defining and Measuring Information Flows within Tweets during Scientific Conferences==
https://ceur-ws.org/Vol-718/paper_04.pdf
Citation Analysis in Twitter:
Approaches for Defining and Measuring Information
Flows within Tweets during Scientific Conferences
Katrin Weller1 , Evelyn Dröge 1 and Cornelius Puschmann 2 ,
Heinrich-Heine-University Düsseldorf,
1
Dept. of Information Science & 2 Dept. for English Language and Linguistics,
Universitätsstr. 1, 40225 Düsseldorf, Germany
{katrin.weller, evelyn.droege, cornelius.puschmann}@uni-duesseldorf.de
Abstract. This paper investigates Twitter usage in scientific contexts, particu-
larly the use of Twitter during scientific conferences. It proposes a methodology
for capturing and analyzing citations/references in Twitter. First results are
presented based on the analysis of tweets gathered for two conference hashtags.
Keywords: Twitter, microblogging, tweets, citation analysis, informetrics.
1 Introduction
With its enormous gain in popularity, the microblogging service Twitter has already
become the subject of different scientific studies. [1] were among the first to
investigate why and how people use Twitter: “From our analysis, we find that the
main types of user intentions are: daily chatter, conversations, sharing information
and reporting news. Furthermore, users play different roles of information source,
friends or information seeker in different communities”. Studies from different fields
of research exist that focus on specific application areas, for example Twitter in
politics and elections [2], in organizational informal communication scenarios [3] or
during natural disasters [4]. Within this paper, we investigate Twitter usage in
scientific contexts and consider Twitter as a means for scientific communication. The
scientific use of Twitter has received some attention in previous work, e.g. [5], [6],
[7], [8]. Our paper suggests refinements of analyzing datasets based on tweets
collected during scientific conferences and present our results from applying novel
forms of intellectual tweet content analyses. Our overall aim is to better understand
how scientists use Twitter and whether traditional patterns of scientific
communication are being mapped to microblog communications or whether entirely
new practices emerge. Therefore we consider information flows as an aspect of
citation analysis within scientific Twitter communication. Scientific communication
in its classical form of publications and citations has long been a subject to analyses
in the fields of scientometrics and informetrics. Informetric citation analysis
distinguishes citations from references [9]: A citation is a formal mention of another
work in a scientific publication – viewed from the cited work‟s perspective. A
1
· #MSM2011 · 1st Workshop on Making Sense of Microposts ·
reference is the same mention of a work but viewed from the citing work‟s
perspective (typically in form of a reference section in a publication). Thus, citations
and references are two sides of the same coin 1. Our paper investigates whether and
how similar information flows exist in microblog communications by comparing two
types of citations on Twitter: URLs pointing to external resources and retweets (RTs)
that cite other users‟ tweets (a more detailed definition will be given in section 2.2).
For both types, we propose methodological approaches to performing citation analysis
on conference tweets data and present first results for these approaches based on a test
dataset collected via the hashtags of two scientific conferences. This paper should
thus primarily be viewed as exploratory research in the field of informetrics for
microblogging. It may provide a basis for future work on developing novel
informetric indicators or for the development of applications that make use of these
indicators, e.g. for identifying and ranking popular tweets, popular twitterers or
external resources, as well as for displaying user networks based on co-ciation or
bibliographic coupling. This work thus also relates to webometrics [10], a sub-
discipline of informetrics that discusses metrics for information exchange and com-
munication on the Web. Recently, new Web 2.0 tools that enable novel forms of
social interaction have brought about a range of new aspects that can be measured and
evaluated (e.g. relating to access and usage, Web publication behavior, user
interrelations). [10] explains that measuring Web 2.0 services offers new ways for
data mining; it can help to gain insights to “patterns such as consumer reactions to
products or world events”. [11] provide an overview on Web 2.0 services (including
microblogging) that may be of interest for new scientometric indicators by measuring
publication impact based on social mentions .
2 Identification of Scientific Microblogging Activities
Twitter is a tool which is not dedicated to one particular application scenario and thus
includes users with various backgrounds and different motivations. It is difficult to
identify scientific tweets or twitterers for analyses.2 In the next section we will discuss
the challenges of gathering data about scientific Twitter usage in order to explain why
our datasets are purely based on hashtags from scientific conferences and thus to
indicate some limitations of our current approach.
2.1 Basic Problems in Identifying Scientific Microposts
Currently, there are no reliable statistics about how many scientists use Twitter (and
more specifically, how many of them do so for science-related communication).
Empirical studies (quantitative and qualitative designs) that investigate scientists‟
motivations for using Twitter are still missing. Presumable reasons for using Twitter
might be timely access to novel information sources and spontaneous creation of
1 We will use „citation‟ as the broader term for both citations and references.
2 In a fundamental consideration one may furthermore discuss the proper definition for what
exactly counts as a scientist or a scientific publication, but this is not a focus of our work.
2
· #MSM2011 · 1st Workshop on Making Sense of Microposts ·
networks based on shared interests (e.g. via hashtags), as well as general benefits of
informal communication as identified by [3]. There is also no general definition of
scientific tweeting. It may for example refer to the following aspects:
Any tweet with scientific content or linking to scientific content: The scientific
Twitter data could be a set of tweets with actual scientific contents. This, however,
is almost impossible to achieve, as it would require either manual identification of
tweet contents or elaborated computer-linguistic automated methods as well as an
elaborated definition for „scientific contents‟. Another interesting subset of Twitter
is the number of tweets that include links to purely peer-reviewed scientific
publications on the Web [11]. Yet, currently tweets with links to scientific
publications are also difficult to collect automatically.
Any tweet published by a scientist: Analyses of scientific microblogging may be
entirely based on its users. Such approaches are frequently applied in analyses of
scientific blogging, while the definition of „scientist‟ in this context may be narrow
(only including members of universities) or broad (including also, for example,
teachers and science journalists). In analyzing Twitter based on users, one always
depends on the biographical information provided by the twitterers. Furthermore, a
selection of users will have to be made manually. [12] have for example manually
identified 28 twittering scientists (using a snowball system) to analyze their
citation behavior. [13] has identified twitterers with academic background by
examining the list of followers of the Chronicle of Higher Education‟s Twitter
account. To our awareness, there are so far no studies that analyze Twitter accounts
belonging to scientific groups or institutions.
Any tweet with a science-related hashtag: Finally, one may identify scientific
tweets based on hashtags. In still rather rare cases, scientists announce particular
hashtags for their projects or topics of interest. One example is the hashtag
“#altmetrics” which is introduced by [14] for work on measuring scholarly impact
on the Web. More frequently, we find specific hashtags for scientific conferences,
some of them officially proposed by the organizers (e.g. “#websci10”). So far,
most studies on scientific microblogging have used datasets collected via con-
ference hashtags. For example, [6] and [7] have gathered sets of conference tweets
to perform automatic analyses on measures such as the number of tweets, the most
active twitterers and the dynamics of the conference. [15] are developing automatic
methods for extracting semantic information from conference tweets. [5] and [8]
have performed manual/intellectual categorizations of tweet contents. This paper is
the first to focus on Twitter citations in the context of scientific conferences.
2.2 Citation Analysis on Twitter
[12] define Twitter citations as “direct or indirect links from a tweet to a peer-
reviewed scholarly article online” and distinguish first- and second-order citations
based on whether there is an “intermediate webpage between the tweet and target
resource”. In their sample of tweets collected from 28 academics they discovered that
of all tweets including an URL, 6% fit into their definition of twitter citations, i.e.
they linked directly or via an intermediate page (like a blog post) to a peer-reviewed
3
· #MSM2011 · 1st Workshop on Making Sense of Microposts ·
article. We suggest that linking to a peer-reviewed publication is only one possible
dimension of citing with Twitter and want to discuss the following alternatives:
All URLs included in tweets may be counted as a form of reference. Analyses may
focus on the types of resources that are referenced in URLs. URLs in tweets act as
external citations (where the tweet includes a reference and the external source
receives a citation).
Retweets can be interpreted as a form of inter-Twitter citations (internal citations).
A user who retweets another one publishes a reference, the retweeted user gets a
citation. In general, users retweet for different reasons like information diffusion or
use retweets as a “means of participating in a diffuse conversation” [16]. Yet,
retweet analyses are not easy to perform, due to the lack of format standardization.
@mentions of usernames within tweets also sometimes resemble references, e.g. in
tweets like “Just read an interesting paper by @sampleuser”. Yet, they can
currently not be automatically distinguished from other @messages and will thus
have to be excluded from current analyses.
In the following section, we will exemplarily analyze and compare some test sets of
hashtag-based conference tweets with regard to the first two types of Twitter citations,
namely URLs in tweets (external citations) and retweets (internal citations).
2.3 Data Collection
For our study we have adapted the conference hashtag principle 3 to gather a collection
of tweets. During our previous work [5] we collected tweets from four scientific
conferences; we selected two smaller conferences (<500 participants) and two major
conferences (>1.000 participants), with one small and one larger conference on topics
from (digital) humanities and one small and one larger conference in the field of
computer sciences. In [5] we performed intellectual analyses of tweets in these
conference datasets. In this paper we now continue this work and perform the
additional manual analysis of URLs included in tweets. For this purpose we have
chosen the two major conferences investigated in [5], namely the World Wide Web
Conference 2010 (WWW 2010, hashtag #www2010) and the Modern Language
Association Conference 2009 (MLA 2009, hashtag #mla09), as we expected to find
discipline-specific differences there. Table 1 presents an overview of the key infor-
mation about the selected conferences and their respective hashtags. It is necessary to
point out that this approach inevitably leads to loss of data: there may be tweets about
the conferences without these particular hashtags or with misspelled hashtags (e.g.
#www10). While typical misspellings may be considered for data collect ion, tags
without any referencing hashtag cannot be collected for events like conferences. As
we could not guarantee to capture all spelling variants for the conferences in our
dataset 4, we deliberately concentrated on the main hashtag for each conference in
order to achieve uniform preconditions for each set. For the same reason, we did not
3
We intend to broaden the approach and want to analyze and compare additional datasets based
on identified scientific twitterers in future work.
4 This is mainly due to limitations to retrieve tweets older than a few days via the Twitter API,
as there were no tweet archives available for all possible spelling variants.
4
· #MSM2011 · 1st Workshop on Making Sense of Microposts ·
include hashtags for associated or co-located events (e.g. #websci10 for the Web
Science Conference co-located with WWW 2010). Tweets were collected for a period
starting two weeks before and ending two weeks after the conference (Table 1). 5
Table 1. The test dataset for tweets with conference hashtags #mla09 and #www2010.
Hashtag #www2010 #mla09
Conference World Wide Web M odern Language
Conference (WWW Association Conference
2010) (M LA 2009)
Conference location Raleigh, NC, USA Philadelphia, PA, USA
Conference dates 26.-30. April 2010 27.-30. December 2009
Discipline Computer science Linguistics, literature,
(digital humanities)
No. of tweets from two weeks 3,358 1,929
before until two weeks after the [during period: 13. April [during period: 15. Dec.
conference 2010-14. M ay 2010] 2009-14. January 2010]
Total no. of unique twitterers 903 (∅ 3.72) 369 (∅ 5.23)
(average no. of tweets per twitterer)
Total no. of tweets during actual 2,425 1,206
conference days only [26.-30. April 2010] [27.-30. December 2009]
4 Analysis of URLs in Tweets
Within our two datasets of #www2010 and #mla09 tweets, we identified all tweets
that include an URL as a link to a website6 as an external citation. Within Twitter,
URLs are often shortened with so-called URL shorteners (such as Bit.ly). Shortened
URLs were resolved to create a list of all URLs included in the datasets. Multiple
appearances 7 of exactly the same URLs 8 could be identified and counted (Table 2).
A basic categorization scheme was developed to classify types of websites that the
URLs included in tweets are pointing to. Each URL within the dataset was classified
by hand according to the following scheme:
5 We may now principally analyze data for this entire period or for the actual conference days
only. If not indicated otherwise, all numbers in the following sections refer to the broader
period from two weeks before until two weeks after the conference dates.
6 URLs were detected by the character strings “http://”, “https://” and “www.” (followed by
additional text, not a blank space). Expressions like „Amazon.com‟ or „Twitter.com‟ are
more difficult to detect automatically and were deliberately left out, as one may not definitely
state that these should act as links to Websites, they may also be interpreted as proper names
of companies or products.
7 URLs may appear more than once per dataset. This may in some cases be due to retweets, in
other cases different users may post the same URL independently .
8
As we worked with automatic techniques, only exact character string matches were identified
as being multiple appearances of the same URL. For more precise results and for subsequent
studies, we suggest to also check URLs with different strings pointing to the same resources,
e.g. “http://twapperkeeper.com/hashtag/mla09” and “http://twapperkeeper.com/mla09”.
5
· #MSM2011 · 1st Workshop on Making Sense of Microposts ·
Blog: This category is used for all kinds of blogs and blog posts as well as other
private commentaries on personal websites.
Conference: This category is used for the official conference websites.
Error: If a URL could not be accessed, it was marked with this category.
Media: This category was applied for all types of multimedia data, e.g. photos,
videos, other types of visualizations and graphics.
Press: This refers to non-scientific publications, e.g. articles in online newspapers
or journals (in contrast to category “blog”, websites in this category have to belong
to a journalistic source).
Project: This category is used for (official) websites by projects (e.g. the website of
a research group or of a scientific project) and project results (e.g. a particular tool
or platform).
Publication: This includes scholarly publications, e.g. an article in a scientific
online journal (these may be open access publication or intermediate pages that
link to paid content). In contrast to category press, URLs in this category should
refer to a publication following scientific criteria, i.e. they should be peer-
reviewed, follow scientific guidelines and be published by a scientific journal or
publishing house or be accepted for a scientific conference. 9 The category also
comprises lists of publications, e.g. tables of content from a proceedings volume , a
scientist‟s website with his personal list of publications .
Slides: This category is used for links to presentation slides, either on presentation
sharing platforms like Slideshare, on personal, institutional or conference websites.
Twitter: This category comprises links to subpages of Twitter, e.g. Twitter profiles,
as well as Twitter-related websites such as Twapperkeeper.
Other: Not specified, everything that does not belong to the categories above. In
future work, URLs classified as “Other” should be investigated in more detail in
order to refine the categorization scheme.
A considerable number of all conference tweets in our dataset includes links.
Within the #www2010 set, 39.85% of tweets included URLs, within the #mla09 set
there were 27.22% tweets with at least one URL. Within the total collection of 1,460
URLs from #www2010, 574 unique URLs have been identified. Thus, each URL
appears 2.54 times at an average (for the #mla09 set: 2.77 times).
Table 2. Different ways to count URL citations in conference tweets.
#www2010 #mla09
Number (and %) of tweets including at least one URL 1,338 (39.85%) 525 (27.22%)
Number of total URLs 1,460 551
Number of unique URLs 574 199
Of course, there are highly cited URLs and those that appear only once, resulting in
a left skewed distribution as depicted in Figure 1. For #mla09 120 URLs (60.3% of
9 We are aware that this definition needs refinements and additional qualitative analysis about
different notions of „scholarly publication‟ across scientific disciplines.
6
· #MSM2011 · 1st Workshop on Making Sense of Microposts ·
unique URLs) and for #www2010 312 URLs (54.36% of unique URLs) appear only
once in the dataset. Table 2 sums up the different ways to count URL citations.
Fig. 1. How often do URLs appear in the dataset?
Table 3. M ost popular URLs for # www2010.
URL Frequency Category
http://blog.marcua.net/post/566480920/twitter-papers-at-the-www-2010- 41 Blog
conference
http://www.danah.org/papers/talks/2010/WWW2010.html 35 Publication
http://kmi.tugraz.at/staff/markus/www2010/www2010_roomstream.html 29 T witter
http://xquery.pbworks.com/rtp-meetup 22 Error
http://www.elon.edu/e-web/predictions/futureweb2010/carl_mala 22 Conference
mud_www_keynote.xhtml
http://www.elon.edu/e-web/predictions/futureweb2010/default .xhtml 18 Conference
http://futureweb2010.wordpress.com/schedule/ 16 Conference
http://www.slideshare.net/haewoon/what-is-twitter-a-social-network-or-a- 13 Slides
news-media-3922095
http://events.linkeddata.org/ldow2010/ 12 Conference
http://opengraphprotocol.org/ 12 Project
http://www.websci10.org/program.html 12 Conference
Table 4. M ost popular URLs for # mla09.
URL Frequency Category
http://amandafrench.net/2009/12/30/make-10-louder/ 27 Blog
http://www.briancroxall.net/2009/12/28/the-absent-presence-todays- 23 Blog
faculty/
http://nowviskie.org/2009/monopolies-of-invention/ 22 Blog
http://chronicle.com/article/missing-in-action-at/63276/ 20 Error
http://www.profhacker.com/?p=4448 18 Press
http://www.samplereality.com/2009/11/15/digital-humanities-sessions-at- 18 Blog
the-2009-mla/
http://chronicle.com/blogpost/the-mlathe-digital/19468/ 16 Press
http://www.profhacker.com/2010/01/09/academics-and-social-media- 15 Press
mla09-and-twitter/
http://academhack.outsidethetext.com/home/2010/the-mla-briancroxall- 15 Blog
and-the-non-rise-of-the-digital-humanities/
http://www.samplereality.com/2010/01/02/the-mla-in-tweets/ 15 Blog
Table 3 and 4 list the top ten most frequent URLs from the #www2010 and the
#mla09 dataset. Such analyses could help to identify the most influential conference
7
· #MSM2011 · 1st Workshop on Making Sense of Microposts ·
contents or those conference aspects that receive high attention (particularly if URLs
link to papers or presentation slides presented at the respective conference). In case of
the MLA 2009 conference this will not work directly: all of the top ten 10 URLs refer
to press reports or blog posts about the conference in general.
Fig. 2. Analysis of URL categories for unique URLs and all aggregated URLs.
In a general analysis of URL categories, great differences can be found between
the profiles of the two test datasets (Figure 2). Twitterers during #mla09 had a general
preference of linking to blog posts (27.14 of unique URLs, 40.29% of total URLs are
categorized as “Blog”) and press articles (21.61 unique URLs; 16.7% of total URLs).
They did not link to any presentation slides (0 times category “Slides”) and hardly to
any scientific publications (3 unique URLs). 11 For #www2010, the percentage of links
to publications and slides is clearly higher, but blogs still play an important role.
Furthermore, 14.11% of overall URLs (6.45% of unique URLs) link to conference-
related websites (e.g. video lectures from the event). At the time of our study (May-
August 2010/ March 2011), a high number of URLs were no longer accessible or
could not be identified due to misspellings (category “Error”).12
10 The URL on rank no. 4 had to be classified as “Error” as the URL cannot be opened, but as it
is located at http://chronicle.com it can also be assumed to have been a “Press” link.
11
M ore qualitative research is needed in order to explore these discipline specific behavior. The
majority of researchers in humanities‟ discipline may not be using presentation slides.
12 The process of identifying URLs and resolving shortened URLs is error prone and can hardly
be re-enacted with consistent results at different points of time.
8
· #MSM2011 · 1st Workshop on Making Sense of Microposts ·
5 Analysis of Retweets
While external citations might become useful for detecting highly cited publications,
presentations or projects, analyses of RTs are promising for identifying influential
persons (or those receiving high attention) during a con ference. So far, we have
analyzed retweets with respect to cited and citing persons and to highly cited tweets.
Table 5. Different ways to count retweets (RTs).
#www2010 #mla09
Automatically detected RTs: Number and percentage 1,121 (33.38% 414 (21.46% of
of RTs in entire conference dataset of 3,358) 1,929)
∅ RTs per twitterer (automatically detected RTs, 1.24 1.12
entire conference dataset)
Retweets including at least one URL 530 207
Manually detected RTs: Number and percentage of 1,318 (39.25% 514 (26.65% of
RTs in entire conference dataset of 3,358) 1,929)
Manually detected RTs: Number and percentage of 828 (34.13% of 269 (30.6% of
retweets in subdataset of tweets during actual 2,426) 1,206)
conference days
Counting retweets automatically may lead to some loss of information. Not all RTs
start with the characteristic “RT @user”-label at the beginning of a tweet. Some may
also be indicated with “via @user”, others simply copy a message without standar-
dized identification mark. Within our analyses, we have also manually classified
tweets as retweets. 13 Table 5 shows the different counts for retweets , among them the
different values for retweets that were automatically detected via the “RT @user”-
label and manually identified retweets. We did not yet distinguish simple RTs from
“encapsulated retweets” [16]. There is a slightly higher percentage of retweets during
the WWW 2010 conference than the MLA 09. For both conferences, a significant
number of additional non-standard retweets could be identified manually: of 1,318
manually identified RTs for #www2010 85% have also been detected automatically
(80% for #mla09 retweets). For #www2010, the percentage of RTs is slightly lower
during the actual conference dates compared to the entire dataset with an included pe -
riod before and after the conference; for #mla09 it is slightly higher during the
conference days.
Retweets can help to identify highly cited persons within a network. In future work
we intend to analyze the networks based on retweets more closely. So far, we have
identified the persons who publish the most retweets and the persons who are often
retweeted during a conference (based on automatically identified RTs). Typically,
13 We automatically
counted tweets starting exactly with the string “RT @”; these counts do
not include tweets where a “RT @” appears at other positions within the tweet text. M anually
identified RTs should comprise all tweets that include copied tweets, whether or not they are
labeled “RT @user”. Yet, the manual identification of RTs is not always error-free and de-
pends on the definitions for labeling a tweet as RT. We aimed to include all tweets with “RT
@user”, “via @user” or “@user” at some position in the tweet and/or identical text strings.
9
· #MSM2011 · 1st Workshop on Making Sense of Microposts ·
these are not the same persons within one conference (Table 6): the top 3 persons who
publish retweets are themselves rather rarely retweeted (#mla09: newfacmajority 1
RT received, ryancordell 3, jcmeloni 5; #www2010: laterribleliz and uncpublichealth
have not received any RTs, olgag has received 18). For both conferences, the three
users who received the most retweets do all belong to the top 10 most active users
with the most tweets in the dataset (#www2010: boraz on rank 1 with 173 tweets,
apisanti on rank 7 with 54 tweets, futureweb2010 on rank 2 with 129 tweets; #mla09:
samplereality on rank 1 with 150 tweets, briancroxall on rank 3 with 61 tweets,
nowviskie on rank 9 with 45 tweets). Future work should include qualitative analyses
to find out more about these persons backgrounds and motivations.
Table 6. Top 3 of highly citing and highly cited twitterers during #www2010 and #mla09.
#www2010 #www2010 #mla09 #mla09
RTs given RTs received RTs given RTs received
laterribleliz (46) boraz (85) newfacmajority (25) samplereality (49)
uncpublichealth (42) apisanty (61) ryancordell (20) briancroxall (35)
olgag (30) futureweb2010 (51) jcmeloni (13) nowviskie (33)
Fig. 3. Distribution of given and received retweets for #www2010 and #mla09.
In future, we intend to describe types of users based on the percentage of received
and given RTs. For #mla09, there are 199 pers ons who have published at least one of
the 413 retweets but only 89 persons who have „received‟ at least one of those
retweets 14. Figure 3 shows the distribution of given and received retweets . It is
furthermore possible to identify particular tweets that were highly cited. Within the
manually collected retweets we have identified the most highly cited original tweets.
Table 7 and 8 show the top 3 most cited tweets 15 for #www2010 and #mla09. Most of
the highly cited tweets do also include URLs – thus, external and internal citations are
interwoven in Twitter. For #mla09, the top 5 RTs all include an URL. The URL con-
tained in the most frequent RT is also the most frequent one in Table 4, RT no. 2
includes the URL on rank 5 from Table 4, the URL in RT no. 3 is on rank 3.
14 For #www2010 there are 574 users who have published at least one retweet and 239 who
have received at least one.
15 Here, only those tweets are summed-up that include the same text and refer to the same user.
10
· #MSM2011 · 1st Workshop on Making Sense of Microposts ·
Table 7. Top 3 retweets for #www2010 (manually detected retweets).
Tweet text and ID From User RTs
a delegação brasileira presente na #www2010 acaba de receber a w3cbrasil 24
notícia: a cidade do Rio de Janeiro sediará a Conferência
#WWW2013 (ID: 13206448810)
twitter roomstreams for every conference room at #www2010 can be mstrohm 16
found at http://bit.ly/bRfE69 #302C (ID: 12881760468)
Summary of Twitter papers presented at #www2010 alisohani 11
http://is.gd/bRqBF (ID: 13268676873)
Table 8. Top 3 retweets for #mla09 (manually detected retweets).
Tweet text and ID From User RTs
Hey, guys, I've blogged about "the amplification of scholarly amanda- 18
communication": Twitter, #M LA09, @briancroxall, & such: french
http://bit.ly/7SRgqZ (ID: 7221520139)
New at ProfHacker: “Academics and Social M edia: #mla09 and profhacker 17
Twitter,” by @GeorgeOnline (and a bunch of you):
http://wp.me/pAGUw-19K (ID: 7566711357)
"M onopolies of Invention:" text of my #M LA09 talk on labor & IP nowviskie 16
issues in humanities collaboration: http://is.gd/5Gckz (ID: 7185970970)
6 Conclusion and Outlook
We have shown that scientists use two types of Twitter citations during scientific con-
ferences. Users cite external sources in form of URLs and quote statements within
Twitter via RTs. This is a first indication that citations/references in Twitter do not
exactly serve the same purposes as classical citations/references . Future work should
investigate more closely why users cite something on Twitter and compare the reasons
with those that have been detected for classical citations. Furthermore, both types of
Twitter citations may act as webometric resources: RTs may help to identify the most
popular twitterers; URLs could be counted to measure impact of referenced
publications or presentation slides. Both types appear with similar frequency within
one dataset, but differences could be identified for the behavior of participants from
the two different conferences. Future work will have to show, whether these
differences indicate discipline-specific characteristics. Plans for successive work are
the inclusion of additional conference datasets as well as the creation of datasets
based on scientific twitterers , the analysis of citation patterns over time and the
inclusion of qualitative work (e.g. intense content analyses and interviews with users).
Acknowledgements
Many thanks to Julia Verbina and Parinaz Maghferat for their contributions to data
collection. Thanks to Bernd Klingsporn for advice and support and to Wolfgang G.
11
· #MSM2011 · 1st Workshop on Making Sense of Microposts ·
Stock for critical remarks. Thanks to our anonymous reviewers for helpful ideas.
Financial support from the Heinrich-Heine-University Düsseldorf for the Research
Group “Science and the Internet” is greatly acknowledged.
References
1. Java, A., Song, X., Finin, T., Tseng, B.: Why we twitter: Understanding microblogging
usage and communities. In Proceedings of the 9th WebKDD and 1st SNA-KDD 2007
Workshop on Web M ining and Social Network Analysis at ACM SIGKDD, San Jose,
California, pp. 56--65. New York: ACM (2007)
2. Gaffney, D.: #iranElection: Quantifying Online Activism. In Proceedings of the Web
Science Conference (WebSci10), Raleigh, NC, USA (2010)
3. Zhao, D., Rosson, M . B.: How and why people twitter: The role that microblogging plays in
informal communication at work. In Proceedings of the 2009 SIGCHI International
Conference on Supporting Group Work, pp. 243--252. New York: ACM (2009)
4. Vieweg, S., Hughes, A. L., Starbird, K., Palen, L.: M icroblogging during two natural
hazards events. In CHI 2010 – We are HCI: Conference Proceedings and Extended
Abstracts of the 28th Annual CHI Conference on Human Factors in Computing Systems,
Atlanta, GA, USA, pp. 1079--1088. New York: ACM (2010)
5. Dröge, E., M aghferat, P., Puschmann, C., Verbina, J., Weller, K.: Konferenz-Tweets: Ein
Ansatz zur Analyse der Twitter-Kommunikation bei wissenschaftlichen Konferenzen. In
Proceedings of ISI 2011: Internationales Symposium der Informationswissenschaft 2011,
Hildesheim, Germany, pp. 98--110. Boizenburg: VWH (2011)
6. Ebner, M ., Reinhardt, W.: Social networking in scientific conferences: Twitter as tool for
strengthen a scientific community. In Learning in the Synergy of M ultiple Disciplines. Euro-
pean Conference on Technology Enhanced Learning, Nice, France. Berlin: Springer (2009)
7. Letierce, J., Passant, A., Decker, S., Breslin, J. G.: Understanding how Twitter is used to
spread scientific messages. In Proceedings of the Web Science Conference (WebSci10):
Extending the Frontiers of Society On-Line, Raleigh, NC, USA (2010)
8. Ross, C., Terras, M ., Warwick, C., Welsh, A.: Enabled backchannel: Conference Twitter use
by digital humanists. Journal of Documentation, 67(2), 214--237 (2011)
9. Stock, W.G.: Information Retrieval. M ünchen, Wien: Oldenbourg (2007)
10.Thelwall, M .: Bibliometrics to webometrics. Journal of Information Science, 34(4), 605--
621 (2008)
11.Priem, J., Hemminger, B. M .: Scientometrics 2.0: Toward new metrics of scholarly impact
on the social Web. First M onday, 15(7) (2010)
12.Priem, J., Costello, K. L.: How and why scholars cite on Twitter. In Proceedings of the 73rd
ASIS&T Annual M eeting on Navigating Streams in an Information Ecosystem, Pittsburgh,
PA, USA, Article No. 75. New York, NY: ACM (2010)
13.Young, J. R.: 10 High Fliers on Twitter: On the microblogging service, professors and
administrators find work tips and new ways to monitor the world. The Chronicle of Higher
Education, 31, A10 (April 10, 2009)
14.Priem, J., Taraborelli, D., Groth, P., Neylon, C.: Alt-metrics: A M anifesto. Retrieved
January 13, 2011, from http://altmetrics.org/manifesto (2010)
15.Stankovic, M ., Rowe, M ., Laublet, P.: M apping tweets to conference talks: A goldmine for
semantics. In Proceedings of the Third Social Data on the Web Workshop SDoW2010,
collocated with ISWC2010, Shanghai, China (2010)
16.Boyd, D., Golder, S., Lotan, G.: Tweet, tweet, retweet: Conversational aspects of retweeting
on Twitter. In R. H. Sprague (Ed.), Proceedings of the 43rd Conference on System Sciences
(HICSS 10), Honolulu, Hawaii, USA. Piscataway, NJ: IEEE (2010)
12
· #MSM2011 · 1st Workshop on Making Sense of Microposts ·