O artigo abaixo mostra, por meio do rastreamento de 600 mil mensagens do Tweeter, feitas por participantes do movimento, que tipo de mensagem passavam, com que propósito e para quem.
Os autores descobriram que
as características da comunicação refletiam, nas comunicações individuais, os esforços de mobilização em nível local (convocar os atos, divulgar os locais de concentração e as orientações práticas de ocupação) e, na comunicação não individualizada, a busca por desenvolver quadros narrativos do propósito coletivo do movimento em nível nacional.
Abstract
Social
movements rely in large measure on networked communication technologies
to organize and disseminate information relating to the movements’
objectives. In this work we seek to understand how the goals and needs
of a protest movement are reflected in the geographic patterns of its
communication network, and how these patterns differ from those of
stable political communication. To this end, we examine an online
communication network reconstructed from over 600,000 tweets from a
thirty-six week period covering the birth and maturation of the American
anticapitalist movement, Occupy Wall Street. We find that, compared to a
network of stable domestic political communication, the Occupy Wall
Street network exhibits higher levels of locality and a hub and spoke
structure, in which the majority of non-local attention is allocated to
high-profile locations such as New York, California, and Washington D.C.
Moreover, we observe that information flows across state boundaries are
more likely to contain framing language and references to the media,
while communication among individuals in the same state is more likely
to reference protest action and specific places and times. Tying these
results to social movement theory, we propose that these features
reflect the movement’s efforts to mobilize resources at the local level
and to develop narrative frames that reinforce collective purpose at the
national level.
Citation: Conover MD, Davis
C, Ferrara E, McKelvey K, Menczer F, et al. (2013) The Geospatial
Characteristics of a Social Movement Communication Network. PLoS ONE
8(3):
e55957.
doi:10.1371/journal.pone.0055957
Editor: Yamir Moreno,
University of Zaragoza, Spain
Received: October 18, 2012;
Accepted: January 4, 2013;
Published: March 6, 2013
Copyright:
© 2013 Conover et al. This is an open-access article distributed under
the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided
the original author and source are credited.
Funding: The authors gratefully acknowledge support from the National Science Foundation [
http://www.nsf.gov/] (grant CCF-1101743), DARPA [
http://www.darpa.mil/] (grant W911NF-12-1-0037), and the McDonnell Foundation [
http://www.jsmf.org/].
Any opinions, findings, and conclusions or recommendations expressed in
this material are those of the author(s) and do not necessarily reflect
the views of the funding agencies. The funders had no role in study
design, data collection and analysis, decision to publish, or
preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
One
of the most prominent American political movements of the past thirty
years, Occupy Wall Street (‘Occupy’) is remarkable in the extent to
which social media played a central role in its development and
organization
[1],
[2].
In this study, we examine how the needs and constraints of social
movements are reflected in the geospatial characteristics and
information sharing practices of Twitter users engaged in communication
about the Occupy movement. Specifically, we focus on the geographic
distribution of these users and the ways in which the relationships
among them diverge from those of users contributing to the two most
popular streams for stable political discourse in the United States,
‘Top Conservatives on Twitter’ and ‘Progressives 2.0.’
The
organizing forces underlying successful social movements have been
studied extensively by sociologists and political scientists
[3]–
[7].
From this body of work common themes have emerged, include the problems
of resource mobilization and collective framing, which together
constitute two of the core issues any social movement must address in
order to effect social or political change. Resource mobilization refers
to the process by which a social movement must marshal the financial,
material, and human resources required to sustain its activities
[8].
Collective framing is a process whereby the constituents of a social
movement, through formal or informal processes, come to establish the
narratives, language, and imagery that capture the essential features of
the movement’s purpose and struggle
[9].
Effective framing helps to foster a sense of community and engagement,
and can be a powerful response to countervailing social pressures from
establishment organizations
[10].
Here
we study Occupy Wall Street, a social movement focused on issues
relating to the uneven distribution of wealth, social inequality,
corporate greed, and the regulation of major financial institutions.
Since the first protest on September 17
,
2011, a major feature of the movement has been the long-term physical
occupation of high-visibility encampments, often found in parks, banks,
libraries and foreclosed homes. As a result, the Occupy movement
requires substantial supporting infrastructure, including housing and
sanitation facilities, as well as access to communication technologies.
In spite of this, Occupy has sustained a lasting presence in American
cities including New York City, Oakland, Washington, D.C., and Boston,
which also represent key loci of decision making and protest activity
[1],
[2].
Under the Occupy model, proposals are brought to a vote before a
general assembly, a form of direct democracy in which any participant is
free to comment or vote on any proposal under consideration. The most
prominent among these organizational structures is the New York City
General Assembly, which has been responsible for producing policy and
key narrative frames such as the popular protest slogan, “We are the
99%,” which references the disproportionate concentration of wealth
among the top 1% of the world’s population
[11].
Social
media have played a prominent role in facilitating communication and
coordination throughout the development of the Occupy Wall Street
movement. For example, the first call to action in the Canadian
anticapitalist magazine ‘AdBusters’ used the Twitter ‘hashtag’
#occupywallstreet as one of just ten words featured in a full-page ad.
Ever since, the Twitter platform has been used extensively by movement
participants
[2],
with #ows being one of the hundred most popular hashtags on Twitter for
the year 2011. In addition to Occupy, Twitter has also played a
prominent role in several foreign social movements, most notably in the
Egyptian revolutionary protests of 2011
[12]–
[14].
In
this work, we seek to understand the relationship between the geospatial
dimensions of social movement communication networks and the
organizational pressures facing such movements. To accomplish this, we
use a state-of-the-art location inference technique to model
relationships among users as a weighted directed network of
communication flows between states, in which the weight of each edge
corresponds to the volume of traffic between pairs of locations. Using
this framework we investigate three distinct relationships: attention
allocation and proximity to on-the-ground events, resource mobilization
and localized information sharing, and the role of collective framing in
long-distance communication.
With
respect to the issue of attention allocation, we find that compared to
stable domestic political communication the Occupy Wall Street movement
exhibits very high levels of geographic concentration, with users in New
York, California, and Washington D.C. producing more than half of all
retweeted content. Aside from these hubs, however, we find that the
appeal of content relating to Occupy Wall Street has a
disproportionately local audience. With extended, high profile
encampments and large-scale protest action playing central roles in the
Occupy movement, we propose that this structural feature reflects the
importance of mobilizing human resources at the local level.
Finally,
we report on evidence indicating that the content of communication at
the national level is distinct from the content of communication among
users in the same state. Comparing intrastate versus interstate
communication, we find that the terms most overrepresented in interstate
communication relate to the movement’s core framing language and the
news media, while the terms most overrepresented in local communication
reference physical places, protest action, and specific times. These
results support the hypothesis that local-level communication activity
is driven by the challenge of resource mobilization, while long-distance
communication is more strongly associated with collective framing
processes.
Materials and Methods
Twitter Platform
Twitter is a popular social networking and microblogging site extensively explored in recent literature
[15]–
[21]. Among others, it has been used to study influence and credibility
[22]–
[26], social structure
[27]–
[29] and to monitor users’ sentiment
[30]–
[33]. Twitter users can post
-character messages containing text and hyperlinks, called
tweets,
and interact with one another in a variety of ways. Communication on
Twitter is characterized by directed, non-reciprocal social links that
allow users to subscribe to the stream of content produced by another
user. The content produced by every user an individual follows is
aggregated into a single streaming feed, from which an individual can
selectively rebroadcast content to his or her followers by choosing to
retweet it. In this way, a retweet serves to broaden the potential
audience of a piece of content, and signifies that information has been
transmitted between two individuals. Hashtags, short tokens prepended
with a pound sign (e.g. #taxes or #obama), constitute another important
feature of the platform, and allow the content produced by many
individuals to be aggregated into a custom, topic-specific stream
including all tweets containing a given token.
Data
The analysis described in this article relies on data collected from the Twitter ‘gardenhose’ streaming API between July
, 2011 and March
,
2012 – a nine month period including the birth and maturation of the
Occupy Wall Street movement. The gardenhose provides an approximately
sample of the entire Twitter stream in a machine-readable format.
Gardenhose tweets include useful metadata, among them a unique tweet
identifier, the content of the tweet (including hashtags and
hyperlinks), a timestamp, the username of the account that produced the
tweet, a free text ‘location’ string associated with the originating
user’s profile, and for retweets, the account names of the other users
associated with the tweet. Tweets from geolocation-enabled mobile
devices also report latitude/longitude coordinates, however the
incidence rate of tweets with this data is not enough to be useful as a
feature in general.
To
isolate a representative sample of Occupy Wall Street content we
flagged for collection any tweet containing hashtags associated with the
Occupy movement, including #ows and #occupy{*} (e.g. #occupywallst,
#occupyboston, etc.). To provide a baseline against which to compare our
observations, we also extracted content originating from the two most
popular communication channels associated with stable domestic political
communication, #tcot (Top Conservatives on Twitter) and #p2
(Progressives 2.0). In total, this sampling procedure produced 1,522,415
tweets associated with Occupy Wall Street and 825,262 tweets associated
with domestic political communication. As this analysis is concerned
primarily with information spreading processes we consider only retweet
events from this corpus, resulting in 676,369 retweets among 257,657
users associated with Occupy Wall Street, and 259,703 retweets among
68,049 users associated with stable domestic political communication.
Henceforth, we consider these corpora to constitute representative
samples of retweet interactions among users participating in the streams
of content associated with the Occupy Wall Street movement and stable
domestic political communication in the United States.
Geocoding
To
facilitate a geospatial analysis of communication activity associated
with these content streams we require a high quality method to infer
individual users’ locations. To accomplish this, we rely on
self-reported location strings and the services of a commercial
geocoding API. This technique, popularized in work by Onnela et al.
[34], has been shown to produce high-resolution, high-quality geolocation data in the presence of geographically meaningful input.
A
caveat to this technique, however, is that it relies on raw text
generated by a broad swath of the Twitter population, and so we find
geographically meaningless location descriptors included in the dataset.
To address this issue we rely on an extensive hand-curated blacklist of
popular non-geographical responses such as ‘everywhere’ and ‘the dance
floor’. To produce this list we sorted all location strings by
popularity and reviewed the thousand most popular strings manually,
blacklisting those that did not correspond to geographically meaningful
entities. Drawn from a long tailed distribution, 53% of all tweets in
the data set are associated with a location among the 1,000 most popular
responses, with 27% of all tweets containing one of the top hundred
location strings. From this set of one thousand we blacklisted 161
non-location strings, corresponding to 6% of the tweets associated with
the 1,000 most popular responses.
To improve recall in the presence of novel input, we used a modified version of the Ratcliff-Obershelp algorithm
[35]
to detect fuzzy matches between free text location strings and the
blacklist of popular non-location responses. As a result, because ‘the
dance floor’ is in the set of blacklist responses, strings taking a
slightly modified form, such as ‘on the dance floor,’ will also be
classified as invalid input. The hand-coded blacklist combined with the
Ratcliff-Obershelp fuzzy matching technique resulted in 9% of the
free-text location strings being classified as non-location input.
From
among the remaining responses we submitted location strings to the
Bing.com geocoding API, which returns a best-guess estimate for the
corresponding physical coordinates. This output is hierarchically
formatted to describe the finest level of geographic resolution
available. For example, if a user reports ‘Logan Square, Chicago’ as his
or her location, the Bing API will return information about the likely
zip code, city, state and country associated with that location.
However, if the user reports only ‘USA,’ the information provided by the
API describes only a country-level guess as to the user’s location.
Owing to decreased coverage at the city-level and the proportionately
few users associated with each individual city, we utilize the
state-level location estimates for the geospatial components of this
analysis.
In
total, 68.4% of Occupy Wall Street users reported location strings, and
from these we were able to obtain geolocation estimates for 55.7% of
these accounts. Among this set of users, 60% of the resulting
geolocation estimates included state-level metadata. Response rates were
somewhat diminished for users associated with the stream of domestic
political communication, with 36% of individuals reporting free-text
location strings. Using the procedure described above, we were able to
obtain geolocation estimates for 29.3% of all users in the domestic
political communication stream, 82.4% of which contained state-level
metadata.
Geographic Profile
One
of our goals is to establish a coarse-grained geographic profile for
communication activity associated with the Occupy Wall Street movement.
Formally, for each stream we define an activity distribution across
states as,
, where
is the total number of retweets originating from state
and
is the total number of retweets originating from all states. As
outlined above, we focus on retweets as they correspond to attention
allocation rather than total content production volume.
In
addition to the distribution of activity across individual states we
examine the information sharing relationships among users in different
locations. To accomplish this, we rely on a network representation to
characterize the flow of information on Twitter. Taking users as nodes,
we define a weighted directed network in which an edge with weight
is drawn from node
to
in the event that user
retweets user
times. The intuition underlying this approach is that each retweet
provides evidence suggesting that information produced by user
was evaluated and acted upon by user
.
Combining
the user-level geocode metadata previously described with the network
representation defined here we can induce another network describing the
volume of communication between users in each state. In this network,
nodes represent states, and weighted directed edges are drawn among
them. The weight of the edge from
to
is defined as the sum of the weights among all edges originating from users in state
and terminating in state
.
We note, however, that this induced network must have geolocation
labels for each node in a dyad. In the Occupy Wall Street stream we
identify 143,437 tweets for which both the source and target have
state-level geolocation data and 78,467 likewise restricted tweets in
the stream of stable domestic political communication.
Textual Content
Finally,
we wish to investigate whether the content of tweets with different
geospatial properties serve distinct communication functions. To
accomplish this, we segregate Occupy Wall Street tweets into two
classes:
interstate tweets connect pairs of users in different states, and
intrastate tweets connect users in the same state. We compute the probability of observing a token,
, in a tweet from a given class,
, as
. Comparing these probabilities yields a ratio,
,
a value which is large when a token is more common in intrastate
traffic than interstate traffic and small under the opposite conditions.
Results
Geographic Concentration
Figure 1,
in which states are ordered according to the proportion of stream
activity, shows that content in the Occupy stream is substantially more
geographically concentrated in a few key states compared to domestic
political communication. For example, New York accounts for 30% of the
total retweet activity in the Occupy stream, while the most popular
source for stable domestic political communication, Washington D.C.,
accounts for only 10.7% of the stream’s total volume. As these plots
make clear, the primary locations for on-the-ground Occupy activity are
those places responsible for the majority of widely rebroadcast Occupy
content, with California, New York and Washington D.C. acting as the
source of 53.8% of total retweets.
Figure 2
maps the states where the proportion of activity associated with the
Occupy stream deviates the most from that associated with the stream of
domestic political communication.
We also study the ratio of content production to content consumption by locale.
Figure 3
shows this ratio, defined as the total number of retweets originating
from users in that state divided by the total number of tweets retweeted
by users in that state. This value serves to highlight the extent to
which users in a given location are functioning as content producers or
content consumers. Inspecting this plot, we find that in the Occupy
stream users from just five states produced more content than they
consumed. This stands in contrast to the stream of stable domestic
political communication, in which fourteen states exhibit a ratio
greater than one.
To
highlight the effect of this geospatial concentration on communication
flows between states it is instructive to visualize the structure of
these networks. However, owing to the geographic aggregation process
outlined in section
Geographic Profile both networks are highly
dense, with edges spanning most pairs of states. To address this issue
we utilize a technique known as multiscale backbone extraction
[36],
which is useful for identifying statistically significant edges in
weighted networks, regardless of the absolute value associated with the
weight of that edge. This technique selects for edges with weights
significantly above the expectation given by an analytically defined
probability distribution that models a random allocation of each node’s
strength among its adjacent edges. Parameterized by a confidence level
factor,
,
this technique allows for the selection of statistically significant
edges across all weight scales, a feature that is especially valuable
when working with networks with heterogeneous weight distributions such
as those associated with communication or human mobility.
Applying
this technique to both networks reveals a communication backbone for
the Occupy network that exhibits the highly concentrated hub and spoke
structure described above.
Figure 4
shows that the Occupy Wall Street network is characterized by minimal
state-to-state connectivity, with the majority of statistically
significant traffic flowing to and from New York, California and
Washington D.C. This is in contrast to the communication backbone for
the network of domestic political communication, in which we observe
extensive interactions among many pairs of states.
Localization
In
Figure 5
we present interstate connectivity for each communication network as a
matrix in which the weight of an edge is mapped to a grayscale hue
ranging from white for weak relationships to black for the strongest
relationships. Inspecting these plots, one of the most striking ways in
which the topology of the Occupy Wall Street communication network
departs from that of the domestic political communication network is the
high degree of localization. This is evidenced by the presence of a
strong diagonal in the Occupy Wall Street connectivity matrix, as well
as the significant off-diagonal mass in the domestic political
communication matrix. We find that 40% of Occupy retweets originate and
terminate with users in the same state. In contrast, 11% of retweets
from the domestic political stream exhibit this type of locality, an
increase of more than 350%.
Textual Analysis
To
study the relationship between geography, resource mobilization, and
collective framing, we focus on the content of tweets flowing within and
between states. Restricting our analysis to tokens that account for at
least 0.1% of both the intrastate and interstate tweet text,
Table 1 presents the ten tokens most overrepresented in both intrastate communication as well as interstate communication.
Discussion
The
analysis of interstate connectivity patterns reveals that, relative to
stable domestic political communication, the Occupy network has a highly
localized geospatial structure, with a disproportionately large amount
of traffic being produced and consumed by users in the same state. We
propose that this phenomenon may be related to the issue of resource
mobilization, that is, the process whereby any social movement must
marshall resources such as money, infrastructure and human capital to
further the goals of the movement. In the case of Occupy Wall Street,
such resources are often quite tangible, and include not only tents and
food, but also the participants required to facilitate large-scale
protest action and extended encampments in cities across the country. In
this light, it is easy to understand why such a disproportionately
large fraction of attention is allocated to communication at the local
level.
With
respect to the finding that the majority of widely rebroadcast content
is produced by users in a small number of high profile locations, we
observe that these states represent sites of major encampment and
decision making activity. Despite the fact that all users can contribute
equally to the Occupy stream, it appears that proximity to events on
the ground plays a major role in determining which content receives the
most attention. This is in contrast to the stream of domestic political
communication, in which content from users across the United States is
allocated a significant share of attention. Where the stream of domestic
political communication looks more like a conversation taking place at
the national level, the structure of the Occupy stream is more akin to a
broadcast, with just a few locations playing the role of net content
producers.
Finally,
we propose that interstate communication plays a significant role in
the propagation of narrative imagery associated with collective framing
processes, and that intrastate communication is driven more
predominantly by the pressures of resource mobilization. Looking to the
lists of tokens most overrepresented in each type of traffic (
Table 1),
we find that those more common in interstate communication include
references to core framing language and the news media. This finding
suggests that when users engage in communication across state boundaries
they allocate proportionately higher levels of attention to speech
associated with collective framing processes. In contrast, the tokens
more common in intrastate traffic relate to protest action and specific
times and places. From this we conclude that the content of intrastate
tweets deals much more frequently with rallying the movement’s
participants, a core function of resource mobilization.
The
findings outlined in this paper dovetail nicely with established
literature on social movement theory. However, statistical measures are
limited in the extent to which they can accurately represent nuanced
features of communication, and future work in this domain would benefit
from rigorous qualitative content analysis. Moreover, there remains room
to improve our understanding of how closely the structure of social
media communication mirrors that of other forms of communication. For
example, Mislove, et al. found that the geographical distribution of
Twitter users tends to over-represent populous counties and metropolitan
areas, suggesting that entire rural regions may be significantly
under-represented – with similar findings holding true for ethnicity and
gender as well
[37].
In this respect as well, work of this nature would benefit from deeper
involvement from scholars in the social sciences, and we hope that this
type of interdisciplinary collaboration will become increasingly common.
Acknowledgments
We
would like to thank Alex Rudnick, Jacob Ratkiewicz, Mark Meiss, and
other current and past members of the Truthy group at Indiana University
(cnets.indiana.edu/groups/nan/truthy) for their contributions to the
Truthy Project. Additionally, we would like to thank Fabio Rojas, Brian
Keegan, and Bruno Gonçalves for their constructive, insightful feedback
during the preparation of this manuscript.
Author Contributions
Conceived
and designed the experiments: MDC AF FM EF CD KM. Performed the
experiments: MDC CD. Analyzed the data: MDC AF FM EF CD KM. Wrote the
paper: MDC AF FM EF.
References
- 1.
Chomsky N (2012) Occupy. Zuccotti Park Press.
- 2.
Byrne
J (2012) Occupy the media: Journalism for (and by) the 99 percent. In:
Bryne J, editor, The Occupy Handbook. Little, Brown, 256–264.
- 3.
Myers
D (1994) Communication technology and social movements: Contributions
of computer networks to activism. Social Science Computer Review 12:
250–260.
doi:
10.1177/089443939401200209.
Find this article online
- 4.
Van
Laer J, Van Aelst P (2009) Cyber-protest and civil society: the
internet and action repertoires in social movements. Handbook on
Internet Crime : 230–254.
- 5.
Wray S (1999) On electronic civil disobedience. Peace Review 11: 107–111.
doi:
10.1080/10402659908426237.
Find this article online
- 6.
Bennett
W (2007) Changing citizenship in the digital age. The John D and
Catherine T MacArthur Foundation Series on Digital Media and Learning :
1–24.
- 7.
Bennett
W (2003) Communicating global activism: Strength and vulnerabilities of
networked politics. Information, Communication & Society 6:
143–168.
Find this article online
- 8.
McCarthy
J, Zald M (1977) Resource mobilization and social movements: A partial
theory. American Journal of Sociology: 1212–1241.
- 9.
Benford
R, Snow D (2000) Framing processes and social movements: An overview
and assessment. Annual Review of Sociology : 611–639.
- 10.
Garrett
R (2006) Protest in an information society: A review of literature on
social movements and new ICTs. Information, Communication & Society
9: 202–224.
Find this article online
- 11.
MotherJones website (Accessed 20 Jan 2013). “We Are the 99 Percent” Creators Revealed. URL http://www.motherjones.com/politics/2011/10/we-are-the-99-percent-creators.
- 12.
Choudhary
A, Hendrix W, Lee K, Palsetia D, Liao W (2012) Social media evolution
of the egyptian revolution. Communications of the ACM 55: 74–80.
doi:
10.1145/2160718.2160736.
Find this article online
- 13.
Lotan
G, Graeff E, Ananny M, Gaffney D, Pearce I, et al.. (2011) The
revolutions were tweeted: Information flows during the 2011 tunisian and
egyptian revolutions. International Journal of Communication 5.
- 14.
González-Bailón
S, Borge-Holthoefer J, Rivero A, Moreno Y (2011) The dynamics of
protest recruitment through an online network. Scientific reports 1.
- 15.
Java
A, Song X, Finin T, Tseng B (2007) Why we Twitter: understanding
microblogging usage and communities. In: Proceedings of the 9th WebKDD
and 1st SNA-KDD 2007 workshop on Web mining and social network analysis.
ACM, 56–65.
- 16.
Krishnamurthy
B, Gill P, Arlitt M (2008) A few chirps about Twitter. In: Proceedings
of the 1st workshop on Online social networks. ACM, 19–24.
- 17.
Huberman B, Romero D, Wu F (2009) Social networks that matter: Twitter under the microscope. First Monday 14: 1–5.
doi:
10.2139/ssrn.1313405.
Find this article online
- 18.
Jansen
B, Zhang M, Sobel K, Chowdury A (2009) Twitter power: Tweets as
electronic word of mouth. Journal of the American society for
information science and technology 60: 2169–2188.
doi:
10.1002/asi.21149.
Find this article online
- 19.
Kwak
H, Lee C, Park H, Moon S (2010) What is Twitter, a social network or a
news media? In: Proceedings of the 19th International Conference on
World Wide Web. ACM, 591–600.
- 20.
Boyd
D, Golder S, Lotan G (2010) Tweet, tweet, retweet: Conversational
aspects of retweeting on Twitter. In: Proceedings of the 43rd Hawaii
International Conference on System Sciences. IEEE, 1–10.
- 21.
Wu
S, Hofman J, Mason W, Watts D (2011) Who says what to whom on Twitter.
In: Proceedings of the 20th International Conference on World Wide Web.
ACM, 705–714.
- 22.
Cha
M, Haddadi H, Benevenuto F, Gummadi K (2010) Measuring user influence
in Twitter: The million follower fallacy. In: Proceedings of the 4th
International AAAI Conference on Weblogs and Social Media.
- 23.
Bakshy
E, Hofman J, Mason W, Watts D (2011) Everyone’s an influencer:
quantifying influence on Twitter. In: Proceedings of the 4th ACM
international conference on Web search and data mining. ACM, 65–74.
- 24.
Castillo
C, Mendoza M, Poblete B (2011) Information credibility on Twitter. In:
Proceedings of the 20th International Conference on World Wide Web. ACM,
675–684.
- 25.
Frahm K, Shepelyansky D (2012) Google matrix of Twitter. The European Physical Journal B 85: 1–7.
doi:
10.1140/epjb/e2012-30599-6.
Find this article online
- 26.
Weng
L, Flammini A, Vespignani A, Menczer F (2012) Competition among memes
in a world with limited attention. Scientific Reports 2.
- 27.
Conover
M, Ratkiewicz J, Francisco M, Gonçalves B, Flammini A, et al.. (2011)
Political polarization on Twitter. Proceedings of the 5th International
AAAI Conference on Weblogs and Social Media.
- 28.
Grabowicz
P, Ramasco J, Moro E, Pujol J, Eguiluz V (2012) Social features of
online networks: The strength of intermediary ties in online social
media. PloS one 7: e29358.
doi:
10.1371/journal.pone.0029358.
Find this article online
- 29.
Mathiesen J, Yde P, Jensen MH (2012) Modular networks of word correlations on Twitter. Scientific Reports 2.
- 30.
Tumasjan
A, Sprenger T, Sandner P, Welpe I (2010) Predicting elections with
Twitter: What 140 characters reveal about political sentiment. In:
Proceedings of the 4th International AAAI Conference on Weblogs and
Social Media. 178–185.
- 31.
Bollen J, Mao H, Zeng X (2011) Twitter mood predicts the stock market. Journal of Computational Science 2: 1–8.
doi:
10.1016/j.jocs.2010.12.007.
Find this article online
- 32.
De
Choudhury M, Counts S, Gamon M (2012) Not all moods are created equal!
exploring human emotional states in social media. In: Proceedings of the
6th International AAAI Conference on Weblogs and Social Media.
- 33.
Ciulla
F, Mocanu D, Baronchelli A, Goncalves B, Perra N, et al. (2012) Beating
the news using social media: the case study of american idol. EPJ Data
Science 1: 8.
doi:
10.1140/epjds8.
Find this article online
- 34.
Onnela J, Arbesman S, González M, Barabási A, Christakis N (2011) Geographic constraints on social network groups.
- 35.
Ratcliff J, Metzener D, et al. (1988) Pattern matching: The gestalt approach. Dr Dobb’s Journal 7: 46.
Find this article online
- 36.
Serrano
M, Boguñá M, Vespignani A (2009) Extracting the multiscale backbone of
complex weighted networks. Proceedings of the National Academy of
Sciences 106: 6483–6488.
doi:
10.1073/pnas.0808904106.
Find this article online
- 37.
Mislove
A, Lehmann S, Ahn Y, Onnela J, Rosenquist J (2011) Understanding the
demographics of Twitter users. In: Proceedings of the 5th International
AAAI Conference on Weblogs and Social Media
Para seguir o blog e receber postagens atualizadas, use a opção "seguir", ao lado.
