Menu
Home
Log in / Register
 
Home arrow Computer Science arrow Social Informatics
< Prev   CONTENTS   Next >

3.1 Diffusion Network Structure

To understand the diffusion dynamics of the campaign on Twitter we analyze the retweet-network of the tweets from the dataset. We build a directed graph whose nodes are users and the edges represent users who retweeted tweets from other users (see Figure 1). The weight of each edge indicates the number of retweets between the corresponding adjacent nodes. The resulting graph consists of 3,929 nodes and 6,459 edges. The clustering coefficient of the graph is 0,029.

Then, we apply the Louvain Method[4], a community detection algorithm based on the modularity of the graph, and we identify 61 communities. For each community, we build a sub-graph with the corresponding nodes and the intercommunity edges. We calculate the clustering coefficient of each sub-graph and we identify (1) the node with the highest in-degree (HI), (2) if HI participated in the offline #DaTactic2 event (Madrid or Barcelona) and (3) the type of user HI is (citizen platform, personal account, journalist, politician, ngo, religious platform). The results of the communities formed by more than 10 nodes are presented in Table 3. We observe that, in general, communities whose HI participated in the offline event acquire higher levels of clustering. In fact, the average clustering coefficient of those communities is 0.036 (SD=0.023) while the communities whose HI did not participate in the offline event is 0.009 (SD=0.001).

Fig. 1. Retweet graph of #OccupyEP2014. Nodes are sized by in-degree and colored by the clustering algorithm. The graph is drawn using the OpenOrd layout algorithm[10].

Among these last communities, the only ones with comparable levels of clustering are the ones whose HI is related to the Spanish 15M networked social movement (@DRYmadrid and @Partido X ) or NGOs (@attacespana and @AmigosTierraEsp).

3.2 Impact of Images

To assess the value of images on tweets, out of the 2,945 original tweets, we make a distinction between the tweets that contained an image and the ones that did not. Afterwards, for each tweet, we calculate the number of received tweets and we present in Figure 2 the distribution of tweets over the number of retweets they received. We only consider tweets from 0 to 15 received retweets (>92%) to exclude outliers in the visualization. The results prove the expected

Table 3. Communities, with more than 10 nodes, detected through the Louvain method (N=number of nodes; E=number of edges; HI: node with the highest in-degree in the community subgraph; HI P: if HI participated in the offline event; HI category: category assigned to HI; C c: clustering coefficient of the community subgraph). The communities whose HI participated in the offline event are bolded.

Id

N

E

HI

HI P

HI category

C c

1

387

386

@AsambleaVirtuaI

no

citizen platform

0

2

345

537

@jaazcona

yes

personal

0.057

3

334

410

@lidiaucher

yes

personal

0.035

4

294

294

@otromundoesposi

yes

personal

0.029

5

263

267

@itoguille

no

personal

0.007

6

260

381

@fanetin

yes

journalist

0.05

7

211

232

@DRYmadrid

no

citizen platform

0.025

8

188

221

@Partido X

no

politician

0.031

9

180

199

@AlberAG

yes

personal

0.038

10

156

175

@attacespana

no

ngo

0.04

11

137

201

@frmat

yes

personal

0.089

12

117

118

@15MBcn int

yes

citizen platform

0.004

13

117

121

@TheTroikaParty

yes

citizen platform

0.016

14

106

117

@AmigosTierraEsp

no

ngo

0.029

15

92

92

@Lineasdefuga

no

personal

0

16

88

88

@Famelica legion

no

citizen platform

0

17

79

79

@Stop Monsanto

no

citizen platform

0

18

71

72

@elpidiojsilva

no

politician

0

19

65

69

@serg manero

yes

journalist

0.013

20

64

66

@PatriHorrillo

no

journalist

0.025

21

53

52

@JovenesIUCM

no

politician

0

22

49

48

@arqueoleg

no

personal

0

23

44

44

@RazonFe

no

religious platform

0

24

38

40

@CeliaZafra

yes

personal

0.03

25

35

34

@elNota Lebowski

no

personal

0

26

30

29

@Resetgr

no

personal

0

27

12

11

@3Blackhawk

no

personal

0

hypothesis: the tweets with images were more likely to be re-diffused and get viral. In particular, almost half of tweets without images were not retweeted, whereas only 30% of tweets with images received no retweets.

3.3 Mentions to Political Candidates and Potential Allies

#DaTactic2 participants mentioned 27 political candidates on Twitter. 8 of them interacted and replied to the questions and 8 non-mentioned politicians used the hashtag to spread their ideas about the European election. In addition, the participants mentioned potential allies to spread the actions.

Fig. 2. Distribution of tweets by the number of received retweets

The allies were organized in three categories:

60 profiles related to NGOs (institutional profiles),

35 profiles related to media (digital newspapers and blogs)

60 profiles related to journalists

The analysis of the dataset reveals that the mentioned profiles on Twitter had a notable rate of collaboration:

NGOs: 32 profiles were engaged (53%)

Media: 11 profiles were engaged (31%)

Journalist: 15 profiles were engaged (25%)

 
Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
 
Subjects
Accounting
Business & Finance
Communication
Computer Science
Economics
Education
Engineering
Environment
Geography
Health
History
Language & Literature
Law
Management
Marketing
Philosophy
Political science
Psychology
Religion
Sociology
Travel