=Paper=
{{Paper
|id=Vol-1229/dynak2014_paper7
|storemode=property
|title=Conversations on Twitter: Structure, Pace, Balance
|pdfUrl=https://ceur-ws.org/Vol-1229/dynak2014_paper7.pdf
|volume=Vol-1229
|dblpUrl=https://dblp.org/rec/conf/pkdd/GreethamW14
}}
==Conversations on Twitter: Structure, Pace, Balance==
https://ceur-ws.org/Vol-1229/dynak2014_paper7.pdf
Conversations on Twitter: Structure, Pace,
Balance
Danica Vukadinović Greetham1 and Jonathan A. Ward2
1
Centre for the Mathematics of Human Behaviour
Department of Mathematics and Statistics
University of Reading, UK
d.v.greetham@reading.ac.uk
2
Department of Applied Mathematics
University of Leeds, UK
j.a.ward@leeds.ac.uk
Abstract. Twitter is both a micro-blogging service and a platform for
public conversation. Direct conversation is facilitated in Twitter through
the use of @’s (mentions) and replies. While the conversational element
of Twitter is of particular interest to the marketing sector, relatively few
data-mining studies have focused on this area. We analyse conversations
associated with reciprocated mentions that take place in a data-set con-
sisting of approximately 4 million tweets collected over a period of 28
days that contain at least one mention. We ignore tweet content and
instead use the mention network structure and its dynamical properties
to identify and characterise Twitter conversations between pairs of users
and within larger groups. We consider conversational balance, meaning
the fraction of content contributed by each party. The goal of this work
is to draw out some of the mechanisms driving conversation in Twitter,
with the potential aim of developing conversational models.
Keywords: Twitter mentions networks, conversations models, maximal
cliques
1 Introduction
The rapid uptake of online social media, combined with consumer behavioural
changes around television and news broadcasting, has instigated a sea change
in attitudes within the advertising and marketing sectors. A frequently encoun-
tered adage is that “everything is about conversation and not about broadcast-
ing” [10,6]. By facilitating public addressability through the @ sign (so called
‘mentions’) and enabling private messages, Twitter has confirmed their inten-
tion to function as a communication channel as well as a broadcasting tool.
Access to large quantities of data produced by Twitter users has resulted in a
surge of interest from the academic community [20], who have largely focused
on Twitter’s information flow and retweet behaviour, and hence implicitly the
underlying network of ‘followers’ (e.g. [22,21]). While broadcasting short mes-
sages, or micro-blogging, remains an important component of Twitter use, to
our knowledge comparatively little work has addressed the mining of (public)
conversations on a large scale [3,19,14]. Consequently, we focus in this paper
on analysing the network of communication patterns resulting from mentions in
Twitter.
Although it may not always be clear, even from message content, what in-
tention a user had in mind when posting—information seeking or information
sharing, broadcasting or conversation—we have tried to specifically extract con-
versations by focusing our data-analysis on reciprocated tweets. Moreover, we
have completely ignored the content of conversations and concentrated on struc-
tural and dynamic properties of the underlying mentions network. Our main
objective was to mine actionable insights that could inform our knowledge of
conversational mechanisms and the frequency/timings of tweets. Our hope is
that empirical observations and quantifiable insights from this analysis could
inform a simple, data driven model of the timing and structure of Twitter con-
versations. One possible application would be for automated recommendations
of conversation trends, as discussed in [3,1].
A large number of registered Twitter accounts are operated by automated
software scripts, known as bots [18]. While such accounts are encouraged for
the purpose of developing applications and services, bots whose functions vio-
late Twitter policy (e.g. spammers) are common. The analysis of conversational
patterns and the development of associated models have potential application
for those trying to develop algorithms that can identify nuisance bots. Further-
more, the identification of groups of Twitter users who, through conversational
behaviour, are particularly influential on a specific topic would be particularly
attractive in the marketing sector. Thus, understanding conversational struc-
ture could impact the design and implementation of social media campaigns
and potentially provide a quantitative comparison between Twitter discourse
and other channels of communication, such as face-to-face, telephone, SMS, fo-
rums or email. In addition, curating and recommending conversational trends,
for both Twitter and more generally in online social media, is crucial for social
networking sites as it is one of the main characteristics of user experience. We
believe that a better understanding of the structure, dynamics and balance of
multi-user conversation is key to improving such automated curation systems.
Ultimately, we hope that studying Twitter conversation can ultimately improve
user experience.
In Section 2, we give an account of previous work in this space. Our results
of pairwise and multiple conversations and the Twitter dataset we used are
presented in Section 3. Finally, in Section 4 we summarise and describe possible
directions of future work.
2 Previous work
The phenomenal uptake of Twitter over the last few years has resulted in a
rapidly growing interest in mining Twitter data and particularly sentiment anal-
ysis of tweets. A recent study analyzing a large amount of Twitter and Face-
2
book data [12] found correlations between friendship/follower relations and pos-
itive/negative moods of Twitter users. Diurnal and seasonal mood rhythms that
are common across di↵erent cultures have also been identified in cross-cultural
Twitter data [5], shedding light on the dynamics of positive and negative a↵ect.
A study of conversations within a sample of 8.5k tweets collected over an hour
long period [9] found that the @ sign appeared in about 30% of the collected sam-
ple, its function was mostly for addressing (as intended) and it was relatively well
reciprocated—around 30% of messages containing an @ were reciprocated within
an hour. The majority of these conversations were short, coherent exchanges be-
tween two people, but longer exchanges did occur, sometimes consisting of up
to 10 people. They found that
“...Tweets with @ signs are more focused on an addressee, more likely
to provide information for others, and more likely to exhort others to do
something—in short, their content is more interactive. ”
Twitter conversations also contain both momentarily salient or ‘peaky’ topics,
signified by increased word-use frequency of specific terms, as well as more ‘per-
sistent conversations’, in which less salient terms recur over longer periods [14].
In addition, words that relate to negative emotions are less persistent [22].
In [3], several algorithms for recommending conversations based on the lengths,
topic and ‘tie-strength’3 of conversations were compared. Their results showed
that the di↵erent uses of Twitter (social vs. informational) had a big influence
on the algorithm’s performance — recommendations based on tie strength were
preferred by social users, whilst those based on topic were preferred by informa-
tional users. Related work considered automated curation of online conversations
to present discussion threads of interest to users in e.g. Facebook and Google+.
[1]. Key to this was the prediction of conversation length around a topic and
re-entry of interlocutors. In another work concerning Twitter conversation [13],
a relatively large corpus and content (topic) analysis of 1.3 million tweets was
used to develop an unsupervised model of dialogue from open-topic data.
In our work we completely ignore content, instead focusing on timing, struc-
ture and balance of conversation between pairs of individuals as well as multi-user
conversations. Our contribution is an attempt to map the structure of Twitter
exchanges over a relatively large dataset, while o↵ering some new methods to
mine conversation data and improve statistical models of dialogue.
3 Analysis
3.1 Data
The Twitter data-set investigated in this paper was collected on our behalf
by Datasift, a certified Twitter partner, allowing us to access the full Twitter
3
Tie-strength is an increasing function of the number of exchanged messages between
two people and the number of messages exchanged between them and their mutual
friends.
3
firehose rather than being rate-limited by the API. The data-set consists of all
UK based4 Twitter users that sent tweets with at least one mention between
8 Dec 2011 and 4 Jan 2012 (28 days in total). In the remainder of the paper,
use of the word ‘tweet’ will specifically mean tweets containing at least one
mention. Mentions are messages that include an @ followed by a username.
Thus if person a puts “@b”, it designates that a is addressing the tweet to b
specifically. Mentions are not private messages and can be read by anyone who
searches for them. A tweet can be addressed to several users simultaneously using
@ repetitively. Any Twitter user can mention any other Twitter user, they don’t
have to be related in any way. Since conversational characteristics are influenced
by many factors, including language, culture, community membership etc., one
has to keep in mind the natural limitations of the results of our analysis.
We preprocessed the data, removing empty mentions and self-addressing5
and created a directed multigraph, or mentions network, containing 3, 614, 705
timestamped arcs (individual mentions) from a total of 819, 081 distinct user-
names, or nodes. Of these distinct usernames, 732, 043 were “receivers”, i.e. to
whom a message was addressed, and 137, 184 were “tweeters”, i.e. people who
tweeted a message with a mention. There were approximately 50k nodes that
appeared both as tweeters and receivers. Note that our graph is a multigraph,
meaning that multiple arcs are allowed between pairs of nodes, each having a
direction and timestamp.
3.2 Conversations
An important feature of both face-to-face conversation [16,15] and computer-
mediated communication [8], is the process of turn-taking. Thus in sequences of
mentions between pairs of users, say a and b, we might expect that sequences
like ABABAB would be more common than say AAABBB, where we use A
to denote that party a mentions party b and likewise B to denote that party b
mentions party a.
To establish if this is the case, we assume the null hypotheses that contri-
butions are independent events with probability PA that party a contributes to
a conversation and thus probability PB = 1 PA that party b contributes. For
a given interaction sequence of length N between parties a and b, we are inter-
ested in the number of occurrences of B following A and vice-versa. We call these
transitions, thus the sequence ABAABBA of length N = 7, has 4 transitions.
Note that we focus on reciprocated interactions, meaning that each party makes
at least one contribution and consequently that there is by default at least one
transition in all interactions that we consider. We call the remaining transitions
the excess transitions. For any sequence of length N , the maximum possible
number of excess transitions is clearly N 2. Under the null hypotheses, excess
4
All Twitter users appearing in our data-set had selected the UK as their location.
5
Self-mentioning was surprisingly common in the data-set: 12,680 di↵erent users cre-
ated a total of 44,319 self-mentions, with the maximum being 5,586 from an auto-
mated service that advertises itself at the end of each tweet.
4
transitions occur with probability PT = 2PA (1 PA ). Since we assume that
transitions are independent, the probability distribution of a given number of
excess transitions is binomial, and thus the expected number is ET = (N 2)PT
with variance VT = (N 2)PT (1 PT ).
To test the null hypothesis, we consider all reciprocated pairwise interaction
sequences in our Twitter data-set. For each sequence having nX contributions
from party X 2 {A, B}, we assume that the probability of party a contributing
is simply nA /(nA + nB ). This does not yield any problematic probabilities (i.e.
0 or 1) since both parties always make at least one contribution.
Each sequence may have a di↵erent number of interactions and a di↵erent
transition probability, but assuming that the pairwise interactions are indepen-
dent, the expectation and variance of the ensemble is simply equal to the sum of
the interaction expectations and variances respectively. Doing this, we find that
the expected number of transitions is 85,390 with a standard deviation of 226.3,
but we observe 88,758 transitions in practice, more than 15 standard deviations
above the expected value. We take this as strong evidence that we can reject
the null hypothesis and thus infer that the data contains a significant level of
turn-taking and hence conversation.
Each sequence of pairwise interactions may constitute a number of di↵erent
conversations, but ascertaining when one conversation ends and another begins
may be an extremely difficult task, especially when the goal is to apply an
automated processes to a large data-set. Instead of using a time-intensive lexical
analysis, we investigate whether we can detect conversations by applying a simple
threshold rule to the time gap between responses, where we assume that a time
gap that is larger than the threshold indicates the start of a new conversation.
This method requires that we can identify a suitable threshold. To achieve
this, we divide each sequence of pairwise interactions up according to a given
threshold, then define distinct conversations to be reciprocated sub-sequences,
i.e. sequences containing a contribution from both parties. Thus the number of
sub-sequences nI is always larger than the number of distinct conversations nC .
In Fig. 1(a) and (b) we plot the mean number of sub-sequences and the mean
number of distinct conversations respectively over a range of threshold values.
The number of distinct conversations nC has a peak value at approximately
9hrs. This peak is expected, since we only count reciprocated interactions as
distinct conversations. Thus small threshold values, which split an interaction
sequence up into a large number of short sub-sequences (see Fig. 1(a)), result in
relatively few distinct conversations because many of the sub-sequences feature
contributions from only one party. High threshold values also result in a small
number of conversations, but this is simply because they do not split the sequence
up into many sub-sequences. Thus the maximum at 9hrs is a natural choice
of threshold and corresponds to one’s intuition that conversations may reflect
diurnal patterns.
The mean and median number of tweets during conversations were 13.09 and
4 respectively, but the distribution was heavy tailed (see Fig. 2).
5
(a) (b)
6 1.35
5.5
5 1.3
nI 4.5 nC
4 1.25
3.5
3 1.2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
T T
Fig. 1: Panel (a): Mean number of subsequences for a range of threshold values.
Panel(b): Mean number of distinct conversations for a range of threshold values.
Note that T , time threshold in hours, is normalised on the x-axis.
4
10
3
10
Count
2
10
1
10
0
10 0 1 2 3
10 10 10 10
Conversation Length
Fig. 2: Distribution of conversation length.
We now consider whether the number of contributions from each party are
similar, or ‘balanced’ within pairwise interactions and conversations. For a given
sequence of tweets, there are two ways to compute balance, we can either con-
sider the ratio of means b = hmax(nA , nB )i/hmin(nA , nB )i or the mean of ra-
tios = hmax(nA , nB )/ min(nA , nB )i. We will use the subscripts ‘I’ and ‘C’
to denote whether these have been calculated for interactions or conversations
respectively. Since we only consider reciprocated interactions, both quantities
are well-defined and we would generally expect b < . For the total number of
interactions between pairs, we find that bI = 2.424 and I = 3.457. Thus on
average, one party contributes around 3 times as much as the other. For the
sub-set of conversations, we find that bC = 1.148 and C = 1.425. These are
much closer to 1, and hence more what we would expect from typical, balanced
conversations. The distribution of conversation contribution ratios is plotted in
Fig. 3(a), which illustrates that conversations are most likely to be balanced,
but some extremely unbalanced conversations do occur. In Fig. 3(b), for each
6
(a) (b)
5
10 150
100
Count
nmax
50
0
10 0 1 2
0
10 10 10 0 50 100 150
Balance nmin
Fig. 3: Panel (a): Distribution of conversation balance. Panel (b): Mean maxi-
mum conversation contribution as a function of minimum contribution.
minimum conversation contribution nmin = 1, 2, 3, . . . , we compute the mean of
the maximum contribution nmax . There is a roughly linear trend (the grey line
is nmax = 1.148nmin + 1), which further illustrates conversational balance.
3.3 Multi-user conversations
By allowing multiple @ signs in one message, a Twitter user could send a tweet
to several recipients simultaneously, facilitating multi-user conversations or mul-
ticasting. Note that because of the 140 character limit there is a physical limit
on how many users each message can be multicast to.
In this part of analysis, our aim is to
– Identify multi-users exchanges;
– Determine how many users typically engage in them;
– Identify their time-frame, pace and how balanced they are.
In addition, are all users equally involved, or do some dominate the discussion?
Are the same people at the heart of di↵erent multi-user conversations? What
are the enablers and inhibitors of conversation flowing in the sense of pauses
between consecutive contributions?
3.4 Identification of multi-users conversations
The reciprocated mentions data represents a directed multi-graph G (where
an edge from A to B implies at least one edge from B to A), thus multi-
user exchanges correspond to strongly-connected6 subgraphs of G with k > 2
participants. We ran a non-recursive version of Tarjan’s algorithm [17,11], as
6
A directed graph is called strongly-connected if there is a path from each vertex in
the graph to every other vertex. This means that for two vertices a and b there is
a path in both directions, i.e. from a to b and also from b to a. Strongly-connected
components of a graph are maximal subgraphs that are strongly-connected.
7
implemented in NetworkX [7], to get a list of the strongly-connected compo-
nents of G. Pairwise conversations were discussed in Section 3.2, so we excluded
all strongly-connected components of size 2 from the present analysis. Each
strongly-connected component of at least three vertices was then transformed
into an undirected multi-graph and we ran the NetworkX implementation of the
modified Bron’s algorithm [2] to find all maximal cliques7 . We then disregarded
all cliques of size two. We found in total 2190 cliques of size 3, 4, 5 and 6. The
total number of users in these cliques was 3275 which is around 20% of users
who reciprocated mentions.
In order to take the time elapsed between consecutive messages into account,
we use the same threshold method explained in subsection 3.2, this time demand-
ing for an exchange to be a“conversation” that there is a contribution from all
parties and got relatively similar results (see Fig 4). The number of exchanges
which had contribution from all parties was at peak around 9 and 11 hours. We
took a threshold of 9 hours which gave us 334 multiuser conversations of sizes
3, 4, and 5 (see Fig 5a).
(a) (b)
Fig. 4: Panel (a): Mean number of subsequences for a range of threshold values.
Panel (b): Mean number of distinct conversations for a range of threshold values
(threshold T in hours).
Most users (out of 646) in our dataset were involved in just one multi-user
conversation, but a small number were involved in multiple conversations. The
users’ involvement in multi-user conversation is illustrated in Fig 5b.
When examining the time-frame of multi-user exchanges, we found that the
correlation coefficient between the total number of exchanges between clique
members and the average di↵erence between consecutive exchanges was 0.244
(see Fig 6a). This was not surprising, since we would expect lively conversations
(with lots of exchanged messages) to have a relatively fast pace, in contrast to
a casual exchange of messages with longer di↵erences inside our chosen 9 hour
time-window. The same picture is obtained from looking at the median time
di↵erences between consecutive messages across di↵erent clique sizes (see Fig
7
Maximal cliques are the largest complete subgraphs containing a given node.
8
(a) (b)
Fig. 5: Panel (a): A size of cliques versus a number of instances (log y-axis).
Panel (b):Number of cliques individual users were involved in (log x axis).
6b). We also investigated how balanced multi-user exchanges were, although
(a) (b)
Fig. 6: Panel (a): Average di↵erence in seconds between two consecutive messages
in clique versus total number of exchanges. Panel (b): Histogram of medians of
di↵erences in seconds between two consecutive messages for cliques of size 3,
top, size 4, middle and size 5 bottom.
this situation is more complicated than in the pairwise case.
Firstly, we looked at the di↵erence between the number of tweets received and
sent by individual clique members. For each node, we computed the di↵erence
of their in-degree and out-degree. We summed up the positive values8 and to
normalise, we divided by the total number of exchanged messages. In this way,
we obtained a percentage of ‘unreciprocated’ messages, where reciprocity is not
8
Clearly the number of sent and received messages within a group are equal, thus
summing the di↵erences between in- and out-degree over individual members in the
group is by definition equal to zero.
9
toward a sender but toward a whole group. We show the histograms for the
di↵erent sizes of cliques in Fig. 7a. Across all clique sizes and in most of the
multi-user conversations around 30% messages were unreciprocated. In a small
number of conversations of 3 or 4 users a larger percentage were unreciprocated,
i.e. they were dominated by certain members, but also a large number of cliques
were very balanced (with unreciprocated messages at 0 10%), meaning every
individual received and sent a similar number of tweets.
Finally, we looked at so-called ‘floor-gaining’ [4], i.e. how much input each
user had over the course of a group exchange9 . We compared the out-degree of
each user within a clique, (remember that each clique is a directed multigraph)
with the mean number of edges r = |nE |/|nV |, where nE is the total number of
edges within the clique and nV is the total number of vertices within the clique. In
a ‘round robin’ group conversation, with balanced turn taking, each user would
send out r messages, i.e. be responsible for an equal percentage p = 100r/e of
the total number nE of exchanged messages. For each clique size, we looked at
how many users’ representations were greater than or equal to p, i.e. those users
who ‘dominate’ the conversation. On Fig 7b below, we present the histogram for
a number of dominant users in the cliques of size 3, 4 and 5. This shows that in
(a) (b)
Fig. 7: Panel(a): The percentage of ‘unreciprocated’ messages for cliques of size
3, top, size 4, middle and size 5 bottom. Panel (b): A number of dominant users
in cliques of size 3, top, size 4, middle and size 5 bottom.
most of the cliques of size 3 and 4, one user was responsible for the majority of
communication, whilst in cliques of size five, 2 users were dominant. However in
about 13% of all cliques of size 3 no users dominated, confirming that Twitter
is used for multi-user conversations and not just pairwise conversations.
9
We argue that the action of tweeting in multiuser exchanges can be regarded as floor-
gaining, since tweets with mentions can in principal be read by a wider audience than
the group conversing.
10
4 Conclusions
We looked at conversations in Twitter, based on the underlying structure and
timings in approximately 4 million UK tweets with mentions over a period of
28 days. We structured the data as a multigraph to make use of graph algo-
rithms. We proposed a simple method of identifying conversations between pairs
of users, based on a time-threshold on the time-to-next tweet, and found evi-
dence that a threshold of 9hrs gives a good indication of distinct conversations.
We observed that the conversations detected using this method appeared to be
balanced, meaning that each party involved contributed approximately equally
to the conversation. This was not the case within more general interactions, in
which one agent typically contributed around three times as much as the other.
Although finding cliques in graphs is computationally demanding, because
of the sparsity of interactions patterns within the data-set, extracting multi-
user exchanges was feasible and relatively fast. We were able to find all cliques
within the graph and, using the threshold method, identify conversations for
up to a maximum of 5 users. Most of those exchanges were fast-paced. We also
found that the number of messages in multi-user exchanges was reciprocal to the
average time di↵erence between them. When looking at the balance of multi-user
conversations, we found that most exchanges are dominated by just one or two
users, with some evidence of well-balanced group exchanges in between 3 users.
Regarding the number of received and sent messages by each individual in a
group, we found that some were dominated by one or two users, but also some
were well balanced.
Further work needs to be done using content information to explore how
topics flow through multi-user exchange and if there is any relationship between
time-di↵erences between messages and topic. We hope that the insights gained
from our analysis could help to develop an understanding of the mechanisms and
dynamics of Twitter conversations, with potential scope for generating models
of micro-blogging behaviour.
Acknowledgment
This work is partially funded by the RCUK Digital Economy programme via EP-
SRC grant EP/G065802/1 ‘The Horizon Hub’ and EPSRC MOLTEN EP/I016031/1.
We would like to thank Datasift for the provision of the data analysed, and to
Colin Singleton and Bruno Gonçalves for very useful feedback and comments.
References
1. L. Backstrom, J. Kleinberg, L. Lee, and C. Danescu-Niculescu-Mizil. Character-
izing and curating conversation threads: expansion, focus, volume, re-entry. In
Proceedings of the sixth ACM international conference on Web search and data
mining, WSDM ’13, pages 13–22, New York, NY, USA, 2013. ACM.
11
2. C. Bron and J. Kerbosch. Algorithm 457: finding all cliques of an undirected graph.
Commun. ACM, 16(9):575–577, Sept. 1973.
3. J. Chen, R. Nairn, and E. Chi. Speak little and well: recommending conversations
in online social streams. In Proceedings of the SIGCHI Conference on Human
Factors in Computing Systems, CHI ’11, pages 217–226, New York, NY, USA,
2011. ACM.
4. C. Edelsky. Who’s got the floor? Language in Society, 10:383–421, 1981.
5. S. Golder and M. W. Macy. Diurnal and seasonal mood tracks work sleep and
daylength across diverse cultures. Science, 333:1878–1881, 2011.
6. D. Graham. Twitter: value-added conversation is more important than broadcasts,
rts. memeburn, August 2012.
7. A. Hagberg, D. Schult, and P. Swart. Exploring network structure, dynamics, and
function using networkx. In Proceedings of the 7th Python in Science Conference
(SciPy2008), Passadena, CA USA, pages 11–15, August 2008.
8. S. Herring. Computer-mediated discourse. In D. Tannen, D. Schi↵rin, and
H. Hamilton, editors, Handbook of Discourse Analysis. Oxford: Blackwell, 2001.
9. C. Honeycutt and S. C. Herring. Beyond microblogging: Conversation and col-
laboration via twitter. In Proceedings of the Forty-Second Hawai’i International
Conference on System Sciences, Los Alamitos, 2009.
10. G. Lebo↵. Marketing is a conversation? oh yeah ...about what? The Marketing
Donut, October 2011.
11. E. Nuutila and E. Soisalon-Soinen. On finding the strongly connected components
in a directed graph. Information Processing Letters, 49(1):9–14, 1994.
12. D. Quercia, L. Capra, and J. Crowcroft. The social world of twitter: Topics,
geography, and emotions. In The 6th international AAAI Conference on weblogs
and social media, Dublin, 2012.
13. A. Ritter, C. Cherry, and B. Dolan. Unsupervised modeling of twitter conversa-
tions. In HLT-NAACL 2010, 2010.
14. D. A. Shamma, L. Kennedy, and E. F. Churchill. Peaks and persistence: modeling
the shape of microblog conversations. In Proceedings of the ACM 2011 conference
on Computer supported cooperative work, CSCW ’11, pages 355–358, New York,
NY, USA, 2011. ACM.
15. J. Sidnell. Conversation Analysis: An Introduction. Blackwell, 2010.
16. D. Starkey. Some signals and rules for taking speaking turns in conversations.
Journal of Personality and Social Psychology, 23(2):283–292, 1972.
17. R. E. Tarjan. Depth-first search and linear graph algorithms. SIAM J. Comput.,
1(2):146–160, 1972.
18. K. Thomas, C. Grier, V. Paxson, and D. Song. Suspended accounts in retrospect:
an analysis of twitter spam. In Proceedings of the 2011 ACM SIGCOMM conference
on Internet measurement conference, pages 243–258. ACM, 2011.
19. C. Wang, M. Ye, and B. A. Huberman. From user comments to on-line conver-
sations. In Proceedings of the 18th ACM SIGKDD international conference on
Knowledge discovery and data mining, KDD ’12, pages 244–252, New York, NY,
USA, 2012. ACM.
20. S. Williams, M. Terras, and Warwick. What people study when they study twitter:
Classifying twitter related academic papers. Journal of Documentation, 69, 2013.
21. S. Wu, J. Hofman, W. Mason, and D. Watts. Who says what to whom on twitter.
In WWW, 2011, 2011.
22. S. Wu, C. Tan, J. Kleinberg, and M. Macy. Does bad news go away faster? In 5th
International AAAI Conference on Weblogs and Social Media, 2011, 2011.
12