A Tool for Monitoring of YouTube Content

Intzar Ali Lashari and Uffe Kock Wiil

The Maersk Mc-Kinney Moller Institute, University of Southern Denmark

Campusvej 55, DK-5230 Odense M, Denmark

Keywords:

Social Media, YouTube, Monitoring, Information Visualization, Social Network Analysis and Mining.

Abstract:

The expansion in use of social media has been very signiﬁcant in the past decade. It has become a topic of

interest for many researchers to ﬁnd the most connected people, the most inﬂuential people, etc. in social

media for various purposes. However, collection and monitoring of data in abundance from social media

is difﬁcult. This paper describes a new tool that can collect, monitor, and mine data from YouTube. The

tool is part of a larger framework aimed at monitoring various social media including Facebook, Twitter, and

YouTube. A speciﬁc case focusing on “Islamic Jihad Holy War” demonstrates the features of the tool.

1 INTRODUCTION

A rapidly increasing number of people share informa-

tion with others by using different social media sites

such as Facebook, Twitter, YouTube, LinkedIn, MyS-

pace, etc. A social media site provides a platform to

the users for the social relations based on their respec-

tive interests, relations, and activities which they want

to share with their social circle on the internet. Social

networks can be described as information networks in

which actors are represented by nodes and relations

are represented by edges (Aggarwal, 2011). The con-

text of a social network works as a motivation for the

actors in the social network and the resulting content

generated by the actors and the structure encourages

the participation and subsequently affects the social

relation (Zeng and Wei, 2013). The increase in use

of social media has created an interest in the research

community to analyse and mine the social media data.

However, social network analysis and mining is be-

coming more and more challenging as the generated

content becomes richer and more abundant (Aggar-

wal, 2011).

Social media is used for many different purposes.

Three examples are given here: (1) In countries

like Egypt, Tunisia, and Yemen, rising action plans

such as protests made up of thousands, have been

organized through social media such as Facebook,

YouTube, and Twitter. They used Facebook to sched-

ule the protests, Twitter to coordinate, and YouTube

to tell the world as part of the Arab Spring upris-

ings (2012) (Polymic, 2012). (2) Social media was

extensively used during the East Japan Earthquake

(2011) to share information about the disaster and get-

ting in touch with missing relatives (Telegraph, 2011).

(3) Social media is used increasingly for militant is-

lamist propaganda with the intent to radicalise Mus-

lims (CTA, 2012). Hence, their is an increasing in-

terest in the ability to monitor the social media for

instance to get information about evolving events and

in the interest of public safety.

In this paper, we present a tool for monitoring of

YouTube content. The tool is part of a larger frame-

work aimed at monitoring social media such as Face-

book, Twitter, and YouTube. On YouTube various

types of videos can be uploaded by different users.

Users can watch, like/dislike, and comment upon the

available videos. It is possible to see how many times

a video has been viewed as well as other metrics about

videos and comments. Data changes over time as peo-

ple engage actively on YouTube. Section 2 describes

YouTube in more detail. A set of monitoring metrics

are proposed that can help ﬁnd the most inﬂuencial

people, videos, etc. on YouTube based on social net-

work analysis and mining. Section 3 describes related

work. Section 4 presents the tool and Section 5 fo-

cuses on a speciﬁc case that is used to demonstrate

the features of the tool. Finally, Section 6 concludes

the paper.

2 YOUTUBE

What is the nature of user activity on YouTube? It is

171

Ali Lashari I. and Kock Wiil U..

A Tool for Monitoring of YouTube Content.

DOI: 10.5220/0005068401710178

In Proceedings of the International Conference on Knowledge Management and Information Sharing (KMIS-2014), pages 171-178

ISBN: 978-989-758-050-5

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

an assumption by some scholars (Tannen, 1999) that

in on-line environments the prevalence of anonymity

directly spawns antagonism, and that an increase in

identity information will decrease the communica-

tive hostility. However some scholars hold the op-

posing view that additional identity information, such

as, facial and bodily information, does not guaran-

tee cordial communication among the commentators

(Lange, 2007).

On-line communicative environments like video

sharing sites, blogs, etc. provide a platform for ex-

change of comments. The comments can have a

wide range of reaction, ranging from ecstatic praise

to extreme hate to threat of physical violence. How-

ever, given the fact that on-line interactions take place

in relative anonymity, not all participants may view

certain critical comments as a problem that requires

greater regulatory control, which can be viewed as a

threat to limit participation (Lange, 2007).

Two major types of negative comments can be

distinguished—comments that are hateful or threaten

violence and comments that provide constructive crit-

icism. Hateful comments are often the cause of dis-

couragement of open self-expression on the site. The

driving component in such on-line exchange of com-

ments are the users of video exchange site, blogs, etc.

who participate in the textual communication. Ac-

cording to a previous study, the participants can be

classiﬁed in one of several categories (Lange, 2007):

1. Former Participants. They are those who no

longer post videos, blog posts, etc. but still main-

tain an account.

2. Casual Users. They typically don’t have an as-

sociated account. However, they tend to view

videos or read blogs, etc. when they wish to

search for something while surﬁng the Web, and

are prompted to view a video or read a blog post.

3. Active Participants. They usually have an as-

sociated account, and occasionally participate by

uploading videos, writing posts, and/or by com-

menting on other people’s contributions.

4. Highly-active Participants. These have a more

intense level of activity and participation on the

social sites, spend much more time regularly up-

loading content and maintaining their sites. They

tend to promote their work within and outside of

the content platform.

5. Celebrities. Similar in many respects to the last

category, but are also well-known despite their on-

line presence in the form of YouTube channels,

blogs, FaceBook pages, or Twitter handlers. They

are often in a position to inﬂuence discourse by

the content they upload/create, and other interac-

tions on such sites.

The above categories only provide a description

of the relative levels of participation among the whole

body of users and are not mutually exclusive.

In order to monitor user activity on YouTube, it is

necessary to think in terms of metrics that can quan-

tify users and their activities. Hence, important ques-

tions are: What would we like to monitor? And what

can be monitored given the available YouTube API

(YouTube, 2014)? In terms of videos, it is interesting

to see how inﬂuential they are. This can be measured

by the number of views, the number of likes/dislikes,

and the number of comments. In terms of users, it

is interesting to see how inﬂuential they are and who

they engage with. The former can be determined by

the number of videos they post and how inﬂuential

they are (see above). The latter can be found by ana-

lyzing the social network formed by the activities of

users based on who comments on what videos.

For the purpose of this work, we focus on videos

that are retrieved based on keyword-matching using

YouTube’s API. For each matching video, data about

users who have made comments on that video are col-

lected. The user data collected includes the YouTube

identiﬁers for users who uploaded the video and users

who subsequently commented on the video. This

enables us to generate a social network of YouTube

users from a contextual perspective, where the con-

text is deﬁned by keywords typed in by the person

(investigator or analyst) that wishes to monitor a cer-

tain event, activity, etc.

3 RELATED WORK

Internet-based applications that are build on the tech-

nological and ideological foundation of Web 2.0 al-

lows users to create and exchange the contents of their

interest (Kaplan and Haenlein, 2010). These appli-

cations include Facebook, Twitter, and YouTube to

name a few.

It has been suggested that videos are a very potent

medium for affecting the attitudes and political will

of the intended audience (Farwell, 2010). They can

be used to communicate a message for inﬂuencing

values, culture, attitudes, and opinions. Online video

platforms, like YouTube, provide a very effective and

cheap way to reach mass audiences that would other-

wise be difﬁcult to reach using conventional means.

Terrorist networks has received much attention af-

ter 9/11 (2001). The Al-Qaeda network and related Ji-

hadist organizations have been analyzed with respect

to their Internet based information strategies. Related

KMIS2014-InternationalConferenceonKnowledgeManagementandInformationSharing

172

work in this area has considered online content of Ji-

hadists and their supporters (Conway, 2006) with a

particular emphasis on the content of Jihadi videos

including various types of videos and their impact on

their audiences (Salem et al., 2006), (Kimmage and

Ridolfo, 2007), (Salem et al., 2008).

For example, the HBO documentary called Bagh-

dad ER (Baghdad, 2006), which dealt with the sub-

ject of providing emergency medical care to wounded

US personnel in the battleﬁeld, was re-created by Al-

Qaeda based in Mesopotamia by replacing the origi-

nal soundtrack with their own, and by making an en-

tirely different beginning and ending to show that the

US forces are sustaining loses and being defeated in

combat.

Similarly, in Iraq and Afghanistan, terror groups

use videos to demonstrate their victories over the op-

posing side. Viewer-enthusiasm is gauged by how

quickly the video spreads over the internet and into

news media sites (Farwell, 2010). Because of this un-

conventional approach, they have a greater chance of

getting news coverage via satellite TV which has a

large number of viewers in the Arab world and simi-

lar conﬂict zones.

The main idea is to identify the impact of videos

to the audiences. Funders and policy makers have

shown an increased interest in learning the ways of vi-

olent radicalization (Council of the European Union,

2005). A high proﬁle example is Hussain Osman, one

of the London bombers, who claimed to have been

inﬂuenced by watching videos of the conﬂict in Iraq

along with reading about jihad in an online forum.

In this paper, we create a network of YouTube

video uploaders to investigate the possibilities of rad-

icalization via the Internet speciﬁcally from YouTube

as opposed to analysis of jihadist sites. We aim to ﬁnd

the users that are registered on YouTube and have up-

loaded videos and/or have commented on video con-

tent related to a given context. Hence, our work is

closely related to the work of (Chen et al., 2003; Wen

et al., 2007; Das et al., 2008).

With respect to data collection, analysis, and vi-

sualization, Coplink (Chen et al., 2003) was one of

the ﬁrst systems to successfully address the domain of

criminal network investigation. The system was ﬁrst

deployed at the Tucson Police Department. The sys-

tem collects, combines, and analyzes data from vari-

ous sources and generates overviews of the informa-

tion for the investgators to help them solve cases.

With respect to the idea of context, (Wen et al.,

2007) presents an intelligent information system that

performs an investigation task for detecting frauds.

The authors have contributed by developing two no-

tions: 1. Context and 2. Context-awareness. Re-

lated to context, the paper deﬁnes the term investiga-

tion context and with regard to context-awareness the

investigator can adaptively retrieve data and evaluate

the relevant information for the ongoing investigation.

With respect to collection of user networks from

social media, (Das et al., 2008) focuses on improv-

ing performance in information collection of a so-

cial graph of users’ neighbours in a dynamic social

network. In the study, the author has introduced a

sampling based algorithm for quickly approximating

quantities of interest, the vicinity of a user’s social

graph that explores the variants of correlation across

the sample. The algorithm can be used to rank the

items in the neighbourhood of a user.

Pippal et al. (Pippal et al., 2014) provide a recent

survey of data mining approaches and methods in so-

cial networking sites, including micro-blogging, twit-

ter, YouTube, instagram, blogs, forums, etc. Closely

related work is done by Agarwal and Sureka (Ag-

garwal and Sureka, 2014) where the authors analyse

YouTube metadata for privacy invading and harrass-

ment content. He et al. model user comments on

YouTube videos as a bipartite graph to predict the

popularity of videos and other item on the Web. (He

et al., 2014)

The purpose of our study is to develop a tool for

context-based monitoring of social media data based

on a set of monitoring metrics. To our knowledge, no

one has deﬁned monitoring metrics for YouTube with

the intent to develop a tool to explore and monitor

inﬂuential videos, users, and networks.

4 THE YOUTUBE MONITORING

TOOL

The YouTube monitoring tool is a part of our frame-

work entitled ”Keyword-based Social Network Anal-

ysis Framework” (KSNAF). KSNAF aims at support-

ing collection, monitoring, and mining of social me-

dia data from a contextual perspective. Overall, the

framework must meet the following requirements:

1. The framework enables context-based search. In

the case of YouTube, the framework provides fa-

cilities to collect videos that match a given context

based on keywords deﬁned by the investigator.

2. The framework determines relationships among

the users that are retrieved as the result of a

context-based search. In the case of YouTube, the

framework can build a social network based on

the users’ activities (comments on videos).

3. The framework enables investigators to view

data related to a given set of monitoring met-

AToolforMonitoringofYouTubeContent

173

Figure 1: System architecture of KSNAF.

rics. In the case of YouTube, the framework can

point to the most inﬂuential people, videos, and

(sub)networks.

4. The framework supports social network analysis

related to the generated social network.

5. The framework can export data in formats that

are importable by various social network analysis,

mining, and visualization packages.

An architectural overview of the KSNAF frame-

work is shown in Figure 1. KSNAF includes the fol-

lowing components to meet the above described re-

quirements.

• Comments and Authors Extractor. This com-

ponent is responsible for extracting commenters

and authors with respect to a particular video un-

der investigation.

• Relationship Extractor. This component is re-

sponsible for extracting relationships as ordered

triplets of the form (Author, Commenter, Com-

ment).

• Social Network Mapper. This component is re-

sponsible for mapping the relationships extracted

by the Relationship Extractor component into a

social network. This component contains the co-

ordinating logic that is used in interaction among

other components. At present, the component

uses Algorithm 1 to generate a context-aware so-

cial network.

• Social Network Analyser. This component car-

ries out social network analysis on the networks

generated by the Social Network Mapper com-

ponent. It supports a number of network analy-

sis algorithms, computing various metrics for the

level of nodes, links, and the whole network. It

comes up with results such as what is the most fre-

quent Author-Commenter pattern with respect to

the searched context. Examples of network met-

rics computed by this component include mean

path length, density, and centrality measures (like

degree, betweenness, closeness, etc.). Finally, key

players, weighted link analysis (Memon, 2012)

and determining various clusters in the resulting

social network is supported by this component.

• Analysis Reports. This component generates dif-

ferent reports on the basis of the Social Network

Analyser components’ ﬁndings over a particular

span of time.

• Monitoring Social Network Evolution. This

component is responsible for determining the

changes over time in the structure obtained from

the user-generated activity of writing comments

and/or uploading videos. Hence any changes in

the network are identiﬁed and the network is up-

dated accordingly. These updates are done over

time by the Social Network Analyser component

in a particular search context.

• Social Network Exporter. This component ex-

ports the social network in a variety of formats

to enable social network analysis, mining, and vi-

sualization of third-party software packages. At

present, the component supports comma sepa-

rated values (CSV), XML, and GraphML formats.

In Algorithm 1, V denotes the collection of all

videos, D gives the maximum allowable depth for

search, and d denotes the current search depth, A de-

notes the collection of authors who have commented

on the relevant videos, and ﬁnally G denotes the net-

work graph.

5 CASE STUDY AND RESULTS

In this section, we present a case study to demonstrate

the features of the developed YouTube monitoring

tool. The selected case is related to the uprising

(civil war) in Syria. According to the Danish Se-

curity and Intelligence Service as well as Danish

media (TV2, 2013), (DR, 2014), (Politiken, 2013),

around 90 Muslims have left Denmark to join

the ”holy war” in Syria. In particular, one video

on YouTube is heavily criticized for encouraging

Danish Muslims to join the civil war in Syria.

The video is entitled ”A Danish terrorist in Syria!”

(http://www.youtube.com/watch?v=VTJ1ynW60gU).

This clearly demonstrates the need for a monitoring

tool like the one described in this paper.

KMIS2014-InternationalConferenceonKnowledgeManagementandInformationSharing

174

Data:

Videos V retrieved as result of searching with

context C on YouTube

D is the maximum allowed depth at for

searching

d is the current depth

Result:

Graph G of relationships on YouTube

for each video vx in V do

Retrieve the comment authors A;

Retrieve the comments C;

for each commenter cx in C do

Add an edge that connects A with cx in

Retrieve the videos vx posted by Cx;

if d < D then

Add 1 to d;

Execute the Algorithm 1 with

arguments V x, D, d, G;

else

Return G

end

Algorithm 1: Searching the YouTube graph.

We have used the case study to validate the frame-

work. Basically, the tool is generic and any keyword-

based search of related videos and other associated

data, such as, user IDs, user locations, comments, etc.

can be retrieved from YouTube. The data which we

have collected from YouTube using the available API

(YouTube, 2014) is based on the context “Islamic Ji-

had Holy War.

We provide an analysis of the corresponding so-

cial network that was generated based on the data col-

lected from YouTube. Figure 2 depicts the results of

searching the context and yields the videos that are re-

lated to the context. The resulting videos are then fur-

ther analysed to determine what are the commenters

of the retrieved videos and what are other videos

posed by the commenter. Such analysis is beneﬁcial

to determine the regular commenter on the videos in

the given context, to determine the key players and

sources that positively contributes in uploading the

most viewed or commented videos, and ﬁnally to re-

veal interesting communication patterns like upload-

comment-upload. Figure 3 shows the different social

networks behind each of the videos that is retrieved

against the context search. Table 1 shows the location

of the commenters.

We have run the tool using a Windows 7 PC hav-

ing the following conﬁguration: Intel (R) Core (TM)

2 CPU6600 2.40GHZ and 4.00GB RAM. It took 6

Figure 2: Context-based search of “Islamic Jihad Holy

War”.

Table 1: Location of YouTube commenters.

Country of Users No. of Users

United States (US) 152

Great Britain (GB) 13

Germany (DE) 2

Sweden (SE) 6

Canada (CA) 16

Australia (AU) 6

minutes and 12 seconds to collect the data of author-

ship and generate the commenter clustering. It will

take more time for the collection of data if the depth of

the users’ interconnection is increased. For this case

we have used a depth of 5.

We have used the visualization tool Gephi and the

Yifan Hu graph drawing algorithm (Hu, 2005) which

is efﬁcient and high quality in visualization of rela-

tionships. Figure 4 shows the users as nodes and the

relationships among them, as links, connecting users

who have commented on videos by other users. Al-

though, the networks show different snapshots col-

lected on different dates, the nodes in all the networks

have a one-to-one correspondence among themselves.

Figure 4b-f shows the evolution over time in the

network from March 20 to April 13. Figure 4a shows

the network on April 15. Hence, Figure 4 demon-

Figure 3: Multiple social networks extracted in the context

of the query.

AToolforMonitoringofYouTubeContent

175

(a) (b)

(e) (f)

Figure 4: Network graphs for the extracted YouTube data. (a) shows the entire network (April 15, 2014), (b)–(f) networks

collected from March 20 to April 13, 2014 (newest ﬁrst).

strates that we are able to monitor (track) how the in-

ﬂuence of users and videos change over time. The

highly central users are those whose videos have gen-

erated most comments by other users (e.g., users 47,

92, 49, 86, 22, 91, etc.) as shown in Table 2. It

is interesting to observe that the highly central users

are connected among themselves, i.e., not only does

their own videos generate a large volume of user-

comments (they are popular and inﬂuential) but such

users are commenting on other popular videos as well.

Moreover, Table 3 shows the top most and least pop-

ular users in terms of the number of video views they

have received from other users in the network.

For the collection of data from YouTube we have

used three different API’s provided by YouTube: (1)

data collection based on keyword search, (2) data col-

KMIS2014-InternationalConferenceonKnowledgeManagementandInformationSharing

176

Table 2: Comments on videos by highly central users.

User ID YouTube name No. of comments

47 Jman92854 734

92 meteosurreal 83

49 xHolyCrusader 64

86 Indago55 58

22 Phoebe Igor 55

91 wadeywilson101 48

Table 3: Users with highest and lowest view counts.

User ID Highest

view

count

User ID Lowest

view

count

111 158748 95 3

106 77002 175 12

116 50239 45 33

59 19310 174 35

120 19186 19 38

25 16631 27 38

88 7876 121 43

135 5501 169 44

177 5468 99 44

79 4861 12 46

lection on content authors, and (3) data collection on

author proﬁles. For the information collection of data

of the network we have connected the users informa-

tion on node ID which is the user’s YouTube ID and

related commenters data depending on the depth of

the relations on comments. There is not much varia-

tion in the location of the users from dataset. The ma-

jority of commenters belong to North America (US

and Canada). Figure 4b, the commenter with node ID

1 has claimed that “The Muslims dont need Crusaders

to embrace jihad, jihad is fundamental to Islam” This

is one of the examples of comments obtained from

the dataset based on the mentioned keywords. The

developed tool has the main purpose of supporting

user-deﬁned keyword-based collection of data from

YouTube for monitoring.

6 CONCLUSION

The main goal of this study is to ﬁnd YouTube users

who is uploading videos related to the chosen con-

text and those who have commented on those videos.

Our tool is able to extract data related to YouTube

videos based on keywords chosen by the investi-

gator. The collected data is stored in a relational

database for later operations and queries. The col-

lected data includes user attributes, user comments,

location, timestamps, etc. As far as our tool is con-

cerned, we are able to ﬁnd relations of YouTube video

content and present them as a social network graph

that shows the relationships. In addition, various sta-

tistical metrics of the social network graph are pre-

sented. These metrics show the connectedness and

inﬂuence of users. Our tool is able to detect and up-

date the relationships on a regular basis and show that

on a speciﬁc time who is closely connected to whom

and who is interested in whose content and what con-

tent.

As part of the near term future work, additional

case studies will be made to further validate the

YouTube monitoring tool. Also, we wish to further

investigate how to best visualize the evolution of the

collected metrics and networks over time.

Monitoring tools for Facebook and Twitter are be-

ing developed in parallel to the described tool. They

will be documented in subsequent publications.

REFERENCES

Aggarwal, C. C. (2011). An introduction to social network

data analytics. Springer.

Aggarwal, N. and Sureka, A. (2014). Mining youtube meta-

data for detecting privacy invading harassment and

misdemeanour videos. Master’s thesis, Indraprastha

Institute of Information Technology, Dehli, India.

Baghdad (2006). Baghdad er. http://www.imdb.com/title/

tt0802944/.

Chen, H., Schroeder, J., Hauck, R. V., Ridgeway, L.,

Atabakhsh, H., Gupta, H., Boarman, C., Rasmussen,

K., and Clements, A. W. (2003). Coplink connect:

information and knowledge management for law en-

forcement. Decision Support Systems, 34(3):271–285.

Conway, M. (2006). Terrorism and the internet: New media

new threat? Parliamentary Affairs, 59(2):283–298.

Council of the European Union (2005). The european union

counter-terrorism strategy. Technical report, Council

of the European Union Report.

CTA (2012). Center for terror analysis danish se-

curity and intelligence service. militant islamist

propaganda on facebook and youtube (in dan-

ish). https://www.pet.dk/CTA/∼/media/CTA/

UKLCTAanalyseFacebookogYoutubepdf.ashx.

Das, G., Koudas, N., Papagelis, M., and Puttaswamy, S.

(2008). Efﬁcient sampling of information in social

networks. In Proceedings of the 2008 ACM workshop

on Search in social media, pages 67–74. ACM.

DR (2014). Danish television. 22 young muslisms from

aarhus has left for syria (in danish). http://www.dr.dk/

Nyheder/Ligetil/Dagens fokus/Indland/2014/01/22

unge muslimer fra Aarhus er rejst til Syrien.htm.

Farwell, J. P. (2010). Jihadi video in the war of ideas. Sur-

vival, 52(6):127–150.

AToolforMonitoringofYouTubeContent

177

He, X., Gao, M., Kan, M.-Y., Liu, Y., and Sugiyama, K.

(2014). Predicting the popularity of web 2.0 items

based on user comments. In Proceedings of the 37th

International ACM SIGIR Conference on Research

& Development in Information Retrieval, SIGIR

’14, pages 233–242, New York, NY, USA. ACM.

Hu, Y. (2005). Efﬁcient, high-quality force-directed graph

drawing. Mathematica Journal, 10(1):37–71.

Kaplan, A. M. and Haenlein, M. (2010). Users of the world,

unite! the challenges and opportunities of social me-

dia. Business horizons, 53(1):59–68.

Kimmage, D. and Ridolfo, K. (2007). Iraqi insurgent me-

dia: The war of images and ideas. Radio Free Eu-

rope/Radio Liberty Special Report, page 4.

Lange, P. G. (2007). Commenting on comments: Investigat-

ing responses to antagonism on youtube. In Annual

Conference of the Society for Applied Anthropology.

Retrieved August, volume 29, page 2007. Citeseer.

Memon, B. R. (2012). Identifying important nodes in

weighted covert networks using generalized centrality

measures. In 2012 European Intelligence and Security

Informatics Conference, pages 131–140. IEEE.

Pippal, S., Batra, L., Krishna, A., Gupta, H., and Arora,

K. (2014). Data mining in social networking sites:

A social media mining approach to generate effective

business strategies. International Journal of Innova-

tions & Advancement in Computer Science, 3(2).

Politiken (2013). Danish newspaper. danish muslisms are

encouraged to holy war in syria (in danish). http://

politiken.dk/indland/ECE2047926/danske-muslimer-

opfordres-til-hellig-krig-i-syrien/.

Polymic (2012). Twitter revolution: How the arab

spring was helped by social media. http://

www.policymic.com/articles/10642/twitter-revolution-

how-the-arab-spring-was-helped-by-social-media.

Salem, A., Reid, E., and Chen, H. (2006). Content analysis

of jihadi extremist groups videos. In Intelligence and

Security Informatics, pages 615–620. Springer.

Salem, A., Reid, E., and Chen, H. (2008). Multimedia con-

tent coding and analysis: Unraveling the content of

jihadi extremist groups’ videos. Studies in Conﬂict &

Terrorism, 31(7):605–626.

Tannen, D. (1999). The argument culture: Stopping Amer-

ica’s war of words. Ballantine Books New York.

Telegraph (2011). The telegraph. japan earthquake: how

twitter and facebook helped. http://www.telegraph.

co.uk/technology/twitter/8379101/Japan-earthquake-

how-Twitter-and-Facebook-helped.html.

TV2 (2013). Tv2 news (denmark). terrorism expert:

Danish jihad video is dangerous (in danish).

http://nyhederne.tv2.dk/article.php/id-70767751:

terrorekspert-dansk-jihadvideo-er-farlig.html.

Wen, Z., Zhou, M. X., and Aggarwal, V. (2007). Context-

aware, adaptive information retrieval for investigative

tasks. In Proceedings of the 12th international con-

ference on Intelligent user interfaces, pages 122–131.

ACM.

YouTube (2014). Youtube api v3. https://developers.

google.com/youtube/v3/.

Zeng, X. and Wei, L. (2013). Social ties and user content

generation: Evidence from ﬂickr. Information Sys-

tems Research, 24(1):71–87.

KMIS2014-InternationalConferenceonKnowledgeManagementandInformationSharing

178