Twitter Usage at Conferences

Where are my tweeps?: Twitter Usage at Conferences
Edgardo Vega Ramanujam Parthasarathy Josette Torres
Department of Computer Science Department of Computer Science Department of English
Virginia Tech, Blacksburg, VA, USA Virginia Tech, Blacksburg, VA, USA Virginia Tech, Blacksburg, VA, USA

evega@vt.edu ramanuj@vt.edu josette@vt.edu

ABSTRACT scenarios. As a result, Twitter has changed the question to
“What’s happening?” as it makes more relevance.
The microblogging service Twitter, founded in San Francisco,
California in 2006, has become immensely popular in recent One of the common uses of Twitter is in academic and
times. Apart from serving as a status message update service that professional conferences. Prior research has focused on how
provides the answer to the question "What's happening?" posed individuals have used Twitter as a communication platform. In the
above the update box on the Web site, it has also served as a work of Java et al., boyd et al., and Krishnamurthy et al., they
platform for developing new connections and a medium for studied groups of Twitter users in an attempt to understand why
building conversations. Twitter has found its place as an individuals tweet [2-4]. From this work, the researchers were able
important tool in both professional and academic conference to derive standard metrics for measuring a tweeter’s behavior.
settings. In our study, we aim to analyze how Twitter usage These included measuring the number of tweets, retweets,
patterns change significantly during a conference. We performed followers, and others metrics that we plan to adapt for our own
a series of data analysis tasks on a collection of user tweets study. Some additional relevant findings from these studies are
obtained using the Twitter API from twenty different users for a that the number of followers is not sequential, as previously
five-week period. Our findings revealed that there was substantial assumed [3]. This indicates that community events, such as a
increase in the Twitter usage during the conference week for the conference, may stimulate different behavior. Similarly, in [2],
majority of the users. they found that it was possible to aggregate a community’s tweets
in such a way to discover community intentions. This work
1. INTRODUCTION indicates that communities can have an impact on individual
behavior.

Twitter is a microblogging and social service which allows users The discussion above has led us to study the use of Twitter when
to post 140 character updates, called tweets. Twitter was founded groups gather for an event, such as South by Southwest interactive
in 2006 and has grown to be one of the largest sites on the (SXSWi) conference. These users share the experience of
Internet. What has set Twitter apart in its short but fast attending a conference as a community. The before, during, and
development history is how its users have adapted Twitter for use after behaviors of users in these particular situations - in which we
in ways which were never intended by its developers. What was find the use of Twitter to be extremely interesting - has not been
initially considered only a microblogging platform has quickly studied in depth.
become a tool used by communities to gather around and discuss
events and topics. For example, Twitter has been used as a 2. BACKGROUND
political tool – as it was used in the last presidential campaign, a
tool for gathering public opinions and survey results, a tool to Twitter.com is a popular microblogging service where the users
physically and geographically organizing communities – as it is post status updates, often referred to as “tweets.” While Twitter
used for flash mobs, and many other uses. Each tweet is 140 does have a native web interface, a good percentage of the users
characters or less in length; Twitter’s simple architecture and use third party clients and applications to post updates.
interface makes it appear simplistic. Yet, the variety of uses is
what makes Twitter so appealing yet challenging to study. Twitter 2.1 Conventions and Types
is additionally used as a marketing channel to communicate with
larger communities. For example, TechCrunch, a popular Twitter is being used in many different ways, not merely as just as
technology weblog, publishes links to all their posts on their a status update service. People use Twitter to share
official Twitter stream. This not only gives additional visibility for recommendations, make new friends, track events, and share
the content, but at the same time it helps in initiating pictures, among other uses. The variety of uses is what makes
conversations and comments from fellow Twitter users. Many Twitter appealing and also challenging to study. The tweets, or
other communities and media organizations use Twitter as a updates, posted on Twitter consist of several different types.
broadcasting platform, including CNN, the New York Times, the There are three different types of Twitter users, as described by
Washington Post, and Wired, just to name a few. Java et al., [2] namely the information sources, friends and
Initially, Twitter asked the question, “What are you doing?” with information seekers. The “information sources” might be
the intention of getting the current status update from the user. automated news feeds and tend to have a big follower base, while
However, Mischaud [1] found in his study that more than half of “friends” is a more general category, which includes all the
the tweets he analyzed had nothing to do with the question. common users. “Information seekers” are those users who
People have started using Twitter in many different ways and generally do not post updates, but tend to keep track of updates
posted by other users. While some of the types were present

1

natively, a good number of these features were added to the following distribution in Twitter has a deviation from that of a
service, due to popularity among users. The following are the human social network.
different types of tweets that one could encounter on a typical
Twitter timeline. Twitter facilitates the way in which people use the status message
update feature to ask questions. Often referred to as “social
search,” it involves finding answers with the help of friends in the
1. Regular status updates social graph or even with unknown human resources. There are
2. Replies – Messages in reply to another user (denoted by many motivations like trust, subjective nature of questions, and
@ symbol) human expertise, amongst other reasons for using social search
[6]. Over time, Twitter has shifted from a personal status sharing
3. Retweets – Messages originally posted by another user medium to an information sharing and conversational medium.
and then reposted (typically denoted by “RT”) The way we share and find information has changed with the
4. Followers – People that wish to receive updates from an introduction of microblogging networks, such as Twitter, to the
individual user. Web. A study by Brooks and Churchill [6] involved recording
how users search local news, find shopping deals and use Twitter
5. Following – People that the individual user “follows” for getting recommendations from members of their social graph.
and opts to see their updates. Twitter is like a personal social search service where the
recommendations come from people who we know, unlike
6. Hashtags – A convention of prefixing a # symbol at the
services like Aadvark (vark.com), which provides anonymous
beginning of a keyword, making it easy to filter and
human responses. A more specific utilization in this context can
search tweets.
be seeking time sensitive and location based information.
7. Mentions – Referring to another Twitter user. Denoted
by @ symbol and the Twitter handle. 3.2 Tweet Archives
8. Direct Messaging – A feature used for exchanging
private messages between two users. Commonly Each individual Twitter user sees updates from the people whom
referred to as DM. he or she follows on the personal timeline. The number of updates
9. Messages reposted from other services (location depends on the number of people the user follows; if it is on the
services, RSS feeds, etc.) higher side, the number of tweets that appear on the personal
timeline could be overwhelming and could also cause information
overload. For an individual who has specific tasks and
3. WHY STUDY TWEETS information needs, the process of re-finding will be cumbersome.
The Web interface of Twitter not only archives personal tweets,
Twitter is a huge information repository and provides an ample but also archives tweets from friends in the network. In April
scope for us to look for varied information. It acts as an 2010, the U.S. Library of Congress acquired the entire public
information sharing medium [5], a medium for social search [6], a Twitter archive for preservation and research purposes [11].
phatic communication medium [7], a personal diary, and much
more. Twitter search [8] is turning out to be a valuable tool for
finding information, with one of the biggest advantages being
real-time results. The search stream brings in content from
millions of users and, unlike a normal Web search engine, there is
no issue of indexing latency. This enables the user to get opinions
on any given subject, not only from their personal network of
friends, but also from public users. With location-based filtering,
this will bring in more relevance.
3.1 Information Seeking

The users of Twitter generate over 50 million tweets in a single
day, accounting for over 600 tweets per second [9]. One big task
which might lie in the hands of the user is to find the information
which he/she is looking for amidst thousands of tweets in the
public timeline but, on the other hand, it is undoubtedly a
collective source of intelligence that helps to obtain opinions,
ideas and other relevant information with ease [10]. It serves as a
Fig 1. Representation of the different archives of tweets
medium that enables sharing of information with like-minded
people. Haewoon, et al [5] from KAIST, Korea studied the
influence of Twitter as an information sharing medium by
crawling the entire Twitter site and obtaining a dataset containing The Twitter archive can be viewed from three different
41.7 million user profiles and 106 million tweets, among other perspectives. The first one is the larger subset that contains
information. After ranking Twitter users and analyzing the billions of tweets from all the users of the service. The next level
trending topics and retweets, they have observed that the follower- is the archive of updates from friends in the network. The final

2

one is the personal archive, or in other words, the tweets performed qualitative and quantitative analysis on the anonymized
published by the individual user over a period of time. The level tweet data.
of personal information that can be obtained from this data set
will increase in the outbound manner, with the individual archive 4.1 Data collection
containing more personal references.
There are a few drawbacks in the way of accessing tweet archives, The first step was to choose the conference for which the study
or in other words, re-finding one's own tweets. The current was to be done. We decided on analyzing the tweets from the
Twitter search model only allows data to be searched from tweets SXSW Interactive festival, which took place from March 12-16,
that were published in the past seven days. Although the Twitter 2010 in Austin, Texas. The next step was to choose the users on
profile page has all the tweets, there is no organizational whose tweets the analysis will be performed. We narrowed it
methodology; it is really tedious to look into past tweets. The down to twenty random users who attended the conference, with
Twitter API provides access to a greater number of tweets; there the condition that their tweets were publicly available. The users
are also some external tools and plug-ins that will help to build a were selected based on the Klout influence score [13-
browsable and searchable archive of tweets. Also, Twitter’s direct http://klout.com/kscore/]. The Klout score takes into account the
messaging capability stores private conversations between users; true reach, which is the numerical count of the engaged audience,
these are accessible only to the two parties involved in the the amplification probability, which is ability of the user to initiate
conversation. conversations, and the network score. The scoring algorithm gives
a score ranging between 1 and 100 for any given Twitter user,
wherein a higher score on the scale signifies a stronger influence.
3.3 Conversations and Connections For our study, we selected users whose scores ranged between 24-
84 out of a possible 100 points. This distribution was made to
Twitter also serves as a conversation tool and helps users to make ensure that the users whom we picked for our study had varied
new connections. Honeycutt and Herring [12] analyzed the influence levels among the other twitter users. For instance, a
collaboration and conversation aspects of Twitter by focusing on celebrity user like Bill Gates (@billgates) has a Klout score of 87
user replies and the coherence of exchanges. The @ sign used in of 100, which implies that he is a highly influential user on
the tweets facilitated conversations to a great extent. One other Twitter.
beauty of Twitter is that it develops an ambient intimacy among
the users and serves as a phatic communication medium After deciding on the twenty users, we obtained their tweets using
[7]. Kevin Makice studied this aspect by the Twitter API for a five-week time frame. All tweets published
utilizing “TwitterSpaces,” which are basically large displays by the users, starting two weeks before the conference, during the
streaming tweets from a closely-knit community. It gives a conference week, and two weeks after the conference were
glimpse of the daily happenings of the members and brings in a scraped using the Twitter API. For the twenty users put together,
community feeling. A similar setting is very common in many of we obtained a total of 10,157 tweets. In addition to the actual
the conferences where the tweets from the users attending the tweets, we also recorded additional information such as time the
session are streamed live on a huge screen and this facilitates a tweet was published, device from which the tweet was made,
persistent and mobilizing backchannel environment. McNely [13] presence of hashtags, inclusion of links, etc.
notes that the low barriers to participation, SMS functionality, real
time activity are some of the reasons why microblogging services 4.2 Quantitative Analysis
like Twitter have enhanced the collaborative meaning-making
capacity. The quantitative tweet analysis was performed to obtain statistical
results from the data that we collected. The analysis helped to
Oulasvirta et al [14] discuss the communication genre in understand the change in the usage pattern among the users by
microblogging and how it makes ordinary things visible to others. using the numerical information. The following observations were
The study involved content analysis and qualitative analysis using made by noticing changes in the tweet counts during the five-
a dataset that contained 400000 posts from Jaiku. They focused on week period.
categorizing responses for mundane and non-mundane responses
and emphasized that microblogs are for individuality and serve as • Distribution of tweets
a medium for letting the public know about daily happenings in • Inclusion of hashtags
one's life. Contrastingly, microblogs are being used in many • Distribution of retweets
different scenarios in the past two years; particularly in ways in • Number of replies
which it was not originally designed to be used, such as • Number of mentions
campaigning, customer interaction, and education, among others. • Presence of external links

4. METHODOLOGY 4.3 Qualitative Analysis
Following a grounded theory approach we performed qualitative
This study was focused on finding how Twitter usage varied when analysis, and the 10,157 tweets were manually coded. We
the user was present in a conference setting. We identified users assigned the individual tweets into fifteen specific categories. We
attending a specific conference. By using the Twitter API, we first identified the sentiment of the tweet and marked it as
acquired five weeks’ worth of tweets for those users and positive, negative or neutral. The second attribute was to see if the
tweet was related to the conference or if it was a personal tweet.

3

Each tweet was also tagged with one or more of the following 5. FINDINGS
thirteen additional categories:

1. Emotion: Represents a thought, feeling or behavior. Our findings from the study were two fold. The first one was from
E.g., @jennIRL @GabiKachman Glad your books showed up -- the quantitative analysis and second one was from the qualitative
enjoy! analysis.
5.1 Quantitative Results
2. General Quotes: Reference to a common quotation relevant to
the context or an own quotation. E.g., "It doesn't make economic After performing the quantitative analysis we found a significant
sense to have full-time reviewers" variation in the many of the parameters that we analyzed. Almost
40% replies were made during the third week (conference) and it
3. Observational: A personal observation that mostly results due dropped to less than half of that during the next week. This
to some recent activity. E.g., Twitter Launches A New Dynamic implied that many more conversations happened during the
Homepage (and it's pretty cool) conference time period. The same trend was applicable to retweets
and mentions also. The average number of retweets made by a
user was around 100 during the conference week whereas it was
4. Informational: Provides information that is helpful for other between 40 and 60 during the other four weeks. Over 2000
users. Might be a broadcast to all followers or an individual mentions were made for all the twenty users put together during
response. E.g., @kitson technical issue. working on that. thanks the conference week and it was significantly lower for the other
for the feedback. weeks. The important metric, which we were interesting in
looking at, is the number of tweets made by the user. As indicated
5. Promotion: Used to promote a session or a personal affiliation. by the figure 2, it skyrocketed during the third week, which
E.g., RT @umairh: i'm going to be doing a keynote interview with implied that the user was using Twitter to post updates more
@ev from twitter at sxsw. hope to see many of you there often.

6. Location: Makes reference to a physical location. E.g., at LAX
- 3 hours til the next flight leaves.... I'm getting there.... slowly

7. Location Service: Updated from a location-based service, such
as Foursquare or Gowalla. E.g., Maybe you've heard there is a
rock show here tonight? (@ Stubb's BBQ w/ 115 others)
http://4sq.com/4xtWYc

8. Organization: Mentions an establishment, like a company or a
hotel. Eg., Just arrived at the convention center. Thank goodness
four seasons had umbrellas for guests

9. Recommendations: Giving recommendations about sessions
Fig 2. Distribution of the number of tweets
to attend, places to visit, etc., to other followers. E.g.,
@AmandaMooney You should goto Nick's Crispy Tacos in on Overall, there was significant increase in all the parameters that
Broadway and Polk. There are a ton of great bars there too. we measured and the results are indicated in figure 3

10. Retweets: Re-posting tweets posted by another user. E.g., RT
@joestump: Truth is SimpleGeo is developing AR for NES games.
http://bit.ly/bUDtP2

11. Reporting Quotes: Positing quotations from a session. E.g.,
On web standards: "The minute everyone understands something
is important all progress stops" - @cshirky #sxswi

12. Satisfaction: Displaying a positive tone about a session,
restaurant, etc. E.g., Community funded reporting panel from
@digidave alone made this trip worthwhile for me. #sxswi
Fig 3. Overall quantitative analysis distribution
13. Reference: Making a mention of another Twitter user. E.g.,
Excited to meet @problogger again at SXSWi

4

5.2 Qualitative Results 6. FUTURE WORK

In the qualitative analysis part, we manually tagged over a 10000 Future research on this topic should explore whether elevated
tweets into fifteen different categories. This was performed to see Twitter use is consistent across conferences in different
the context in which the tweet was made. While measuring the disciplines, such as the arts and humanities, government, business,
sentiment we found that majority of the tweets were neutral and and so on. More studies including different conference types
only a few were made either in a positive or negative sentiment should be conducted. Also, interviews with Twitter users who
(fig.4) attend conferences should be conducted so as to gain a better
understanding of individual Twitter usage.

7. ACKNOWLEDGMENTS

Our thanks to Professor Manuel Pérez-Quiñones of the Virginia
Tech Department of Computer Science and Professor Carlos Evia
of the Virginia Tech Department of English for their assistance in
the completion of this research. Thanks to all our classmates from
the Personal Information Management class at Virginia Tech for
providing feedback on our work. We also thank Easy Chair
Coffee shop, Blacksburg for being the perfect environment for all
our team meetings.

8. REFERENCES
Fig 4. Sentiment of the tweets
[1] Mischaud, E. (2007). Twitter: Expressions of the whole self.
An investigation into user appropriation of a web-based
communications platform. London: Media@lse. Retrieved
The major part in the qualitative analysis was to figure out how May 20, 2008 from http://www.lse.ac.uk/collections/
the tweets fit in the thirteen pre-defined categories, which we media@lse/mediaWorkingPapers/MScDissertationSeries/Mi
allotted. We found that 68% of the tweets were observations of sch aud_final.pdf
the happenings. Retweets were next in line and they constituted
10% of the total tweets made. Location based services and [2] Boyd, Danah, Scott Golder, and Gilad Lotan. 2010.
reporting quotes constituted 5% each of the total volume. Figure 5 “Tweet, Tweet, Retweet: Conversational Aspects of
shows the complete distribution of the qualitative analysis. Retweeting on Twitter.” HICSS-43. IEEE: Kauai, HI,
January 6.
[3] Java, A., Song, X., Finin, T., and Tseng, B. 2007.
Why we twitter: understanding microblogging usage
and communities. In SNA-KDD 2007 Workshop on Web
Mining and Social Network Analysis. ACM, 56-65.
[4] Krishnamurthy, B., Gill, P., and Arlitt, M. 2008. A
few chirps about twitter. In Workshop on online Social
Networks. ACM, 19-24.
[5] Kwak, Haewoon, Lee, Changhyun, Park, Hosung, and Moon,
Sue. (2010). What is Twitter, a Social Network or a News
Media?. The 19th World-Wide Web (WWW) Conference.
Raleigh, North Carolina. (Conference paper)
[6] Andrew L Brooks, Elizabeth Churchill, Tune In, Tweet on,
and Twit out: Information snacking on Twitter, CHI 2010
Workshop on Microblogging
[7] Makice, Kevin. (2009). Phatics and the design of community.
It was also worthy to note that only 21.8% of the tweets came
Proceedings of the 27th international conference extended
from the web interface of Twitter. The rest of the tweets were
abstracts on Human factors in computing systems. (pp. 3133-
made from third party clients like Tweetie, Tweetdeck etc. We 3136). (conference paper)
were not able to see if it was published from a mobile device as
the API doesn’t provide complete information on this front. After [8] Gene Golovchinsky, Miles Efron, Making sense of Twitter
analysis of all the data, we concluded that Twitter usage at Search, CHI 2010 Workshop on Microblogging
conferences is significantly more elevated than typical usage. [9] Measuring Tweets,
http://blog.twitter.com/2010/02/measuring-tweets.html

5

[10] Cheong, Mark, Lee Vincent, Integrating web-based [14] Oulasvirta, Antti, Lehtonen, Esko, Kurvinen, Esko, and
intelligence retrieval and decision-making from the twitter Raento, Mika. (2009). Making the ordinary visible in
trends knowledge base, 2009, Proceeding of the 2nd ACM microblogs. Personal and Ubiquitous Computing. Special
workshop on Social web search and mining issue on Social Interaction and Mundane Technologies.
[11] Tweet Preservation , http://blog.twitter.com/2010/04/tweet- (journal article)
preservation.html [15] Spector, A. Z. 1989. Achieving application requirements. In
[12] Honeycutt, C. and S. Herring. Beyond Microblogging: Distributed Systems, S. Mullender, Ed. ACM Press Frontier
Conversation and Collaboration in Twitter. Proc 42nd HICSS, Series. ACM, New York, NY, 19-33. DOI=
IEEE Press (2009). http://doi.acm.org/10.1145/90417.90738.
[13] McNely Brian, Backchannel persistence and collaborative
meaning-making, Proceedings of the 27th ACM international
conference on Design of communication , Bloomington,
Indiana, USA

6

Twitter Usage at Conferences

More Related Content

What's hot (14)

Viewers also liked (6)

Similar to Twitter Usage at Conferences (20)

Recently uploaded (20)

Twitter Usage at Conferences