User recommendations for journalistic websites on Twitter (ICA Presentation 2012, Phoenix)
Hanna Jo vom Hofe, Christian Nuernbergk, Christoph Neuberger
(How to Program) Paul Deitel, Harvey Deitel-Java How to Program, Early Object...
ICA 2012: User recommendations for journalistic websites on Twitter
1. User recommendations for journalistic
websites on Twitter
Hanna Jo vom Hofe, Christian Nuernbergk, Christoph
Neuberger
LMU Munich/University of Muenster
ICA 2012
May 26th 2012
2. Agenda
Introduction: Complementary Relations between Twitter and Journalism
Research Design and Methodology
Findings
Conclusion
3. Complementary Relations between Twitter and Journalism
Professional
Journalism
Promotion of Provision of
news content and story ideas,
news websites sources
(self-promotion, (monitoring, filtering,
branding) reporting)
Twitter Conversation/
Interaction
Editorial Recommendations User Recommendations
of news content of news content
Automated publishing Social Navigation
4. Research Design: LfM Twitter and Journalism Study
1. German Newsroom Survey 2010:
media types: daily and weekly newspapers, general-interest
magazines, supra-regional/national TV/radio, Internet-only news
sites
identification of 157 media outlets/news providers
respondents: editors-in-chief/members of editorial departments
(response rate: 45%, n=70) in May/June 2010
2. Content Analysis of User Recommendations
Detection of all tweets with links to news sites analyzed in the
parallel newsroom survey
Monitoring tool: Backtweets.com web application
5. Methodology: Content Analysis of User Recommendations
354.794 tweets contained a link to one of the 157 sites in April 2010
links pointed to either website domains or specific articles
for each site, the number of in-links was calculated
Systematic sampling and analysis of 1000 tweets
sample inclusion of news sites proportional to their share of in-
links
inclusion of every 1st and 5th hit on each result page for a specific
link search on backtweets.com (27th may 2010) until the previously
calculated proportional share was reached for a site
6. Methodology: Content Analysis of User Recommendations
quantitative analysis of topic area and link type in tweets
destination/reference type (e. g. website, specific article or actors
and events adressed in an article)
evaluation (positive, negative or balanced valence)
coding of n=993 tweets by three coders
inter-coder reliability: 0.94 (Holsti’s coefficient)
units were fully coded if tweets were not published by official editorial
accounts on Twitter (no editorial self-promotion)
exclusion of 186 editorial tweets (19%)
7. RQ 1: What structural patterns do user recommendations
exhibit on Twitter?
How centralized is the distribution of user recommendations to single
news sites?
Do user recommendations reflect the prominence of a news site on the
web (in terms of reach)?
RQ 2: What sort of news gathering, filtering and evaluations
are made transparent through user recommendations?
What kind of topics on news sites are selected for recommendations?
To what extent do user recommendations attach comments to links to
news sites and/or their articles?
8. RQ 1: What structural patterns do user recommendations
exhibit on Twitter?
Professional
Journalism
Promotion of Provision of
news content and story ideas,
news websites sources
(self-promotion, (monitoring, filtering,
branding) reporting)
Twitter Conversation/
Interaction
Editorial Recommendations User Recommendations
19% of news content of news content 81%
Automated Publishing Social Filtering
9. Findings (RQ 1): Incoming Links and Reach of Top 20 News Outlets
In-Links Tweets IVW Visits Rank
Share of
(Tweets) Rank Rank Difference
News Outlet Investigated URL In-Links
in April (April (April (Tweets vs.
in %
2010 2010) 2010) Visits)
Spiegel Online spiegel.de/ 48.794 14 1 2 1
Welt Online welt.de/ 32.792 9 2 4 2
faz.net faz.net/ 23.658 7 3 8 5
Focus Online focus.de/ 17.638 5 4 5 1
tagesschau.de tagesschau.de/ 15.905 5 5 n. a. n. a.
bild.de bild.de/ 14.433 4 6 1 -4
Yahoo! Deutschland de.news.yahoo.com/ 13.681 4 7 n. a. n. a.
Handelsblatt handelsblatt.com/ 12.109 3 8 16 10
Zeit Online zeit.de/ 11.374 3 9 12 5
stern.de stern.de/ 10.278 3 10 10 2
sueddeutsche.de sueddeutsche.de/ 10.220 3 11 6 -3
Financial Times Deutschl. ftd.de/ 9.631 3 12 15 5
n-tv n-tv.de/ 8.322 2 13 7 -4
Der Westen derwesten.de/ 7.328 2 14 18 6
Berliner Morgenpost morgenpost.de/ 5.757 2 15 30 17
taz.de taz.de/ 4.386 1 16 24 10
manager magazin manager-magazin.de/ 3.814 1 17 20 5
Der Tagesspiegel tagesspiegel.de/ 3.675 1 18 22 6
Abacho.de abacho.de/ 3.569 1 19 38 21
Saarbrücker Zeitung saarbruecker-zeitung.de/ 3.506 1 20 46 28
10. Findings (RQ1)
60.000
The 1st top site receives 14%, the 2nd site
Received In-links from twitter
50.000
Spiegel Online receives 9% and the 3rd 7% of all links
40.000 (n=354.794)
Welt Online
30.000 M=2259,83 tweets (SD=5731,11)
Faz.net
20.000
results show centralization to top news sites
Focus Online
10.000
First quintile of the investigated sites share
0
1 51 101 151
82% of all tweets; Top 20 sites receive 74%
Rank
Fig.: Distribution of User Recommendations Power law-distribution
per News site
IVW visits ranking and tweet in-links ranking show a robust correlation
(Spearmans rs=0,736, p<0,01, n=115)
11. RQ 2: What sorts of news gathering, filtering and evaluations
are made transparent through user recommendations?
Professional
Journalism
Promotion of Provision of
news content and story ideas,
news websites sources
(self-promotion, (monitoring, filtering,
branding) reporting)
Twitter Conversation/
Interaction
Editorial Recommendations User Recommendations
of news content of news content
Automated Publishing Social Filtering
12. Findings (RQ 2): Selection of topics in tweeted user
recommendations for Top 20 news sites
Topic area Top 20 news Other news Topics selected for
of tweet/ sites sites recommendation slightly differ
linked article (n=618) (n=186)
between sites
Politics 38% 26%
Economy 15% 16% Recommendations for popular
Culture 4% 10% news sites show preferences
Sports 10% 13% for politics, science and
technology
Media/Net 10% 10%
Science/Technics 8% 4%
Tweets linking to Top 20 sites
Entertainment 5% 6%
comprise only a small amount
Society/Everyday life 5% 10% of societal, everyday life and
Other 6% 5% culture topics
Cramer-V=0,184, p<0,01
13. Findings (RQ 2): Additional Comments and Evaluations
Tweets with links to news sites mostly ignore value judgments:
90% (n=807) of all counted links were not embedded into an evaluative
context.
Only 10% of all recommendations attach comments directed to the
news site, the article of interest, or an event or actors cited/described in
the linked article.
Tweeted value judgments are mainly negative (53%, n=81)
Balanced evaluations remained seldom (9%, n=81)
14. Findings (RQ 2): Additional Comments and Evaluations
Evaluation by subjects in Tweets
Actor or
News site Linked article event in a
(n=8) (n=7) linked article
(n=65)
Negative 25 0 61
Balanced 25 0 8
Positive 50 100 31
Cramer-V=0,362, p<0,01
User recommendations seldomly include evaluations of a linked news site
or an article in general.
More often, tweets discuss or rate actors and events addressed in the
linked articles.
15. Additional Survey Findings
Professional
Journalism
Promotion of Provision of
news content and story ideas,
news websites sources
(self-promotion, (monitoring, filtering,
branding) reporting)
Twitter Conversation/
Interaction
Editorial Recommendations User Recommendations
of news content of news content
Automated Publishing Social Filtering
16. Survey Findings: Recommendations from the editors’ view
According to the survey results from 2010, staff members notice
responses on their own reporting on Twitter
Monitoring: 60% of Top 20 sites members (n=10) and 48% (n=48)
of other interviewees search for user responses on their reporting
Recommendations and site traffic: Almost all respondents (93%, n=42)
estimated traffic amount delivered by Twitter to their news site <10%
In general, survey findings show that Twitter and other social media are
not regularly employed in terms of journalistic research
Limited impact compared to other tools and ways of computer-
assisted reporting (e. g. search-engines, databases)
Staff members most often apply Twitter to check the general opinion
climate (59%, n=58)
17. Conclusion
User recommendations for news sites indicate concentration tendencies
and partially reflect the reach of various news sites.
Only a small subset of websites receives a substantial share of traffic
through Twitter recommendations.
In comparison to other topics, political issues strongly dominate –
recommendation filter and user-led news gathering is biased (in favor of
politics and web news).
Comments are rarely attached to tweets linking to news sites.
Recommendations with comments contain value judgments mainly on
events or actors in linked articles.
18. nuernbergk@ifkw.lmu.de
neuberger@ifkw.lmu.de
hannajo.vomhofe@uni-muenster.de
Neuberger, Christoph/vom Hofe, Hanna Jo/Nuernbergk, Christian (2010):
Twitter und Journalismus. Der Einfluss des "Social Web" auf die Nachrichten.
Düsseldorf: Landesanstalt für Medien Nordrhein-Westfalen (LfM) (=LfM-
Dokumentation Nr. 38).
http://www.lfm-nrw.de/fileadmin/lfm-nrw/Publikationen-
Download/LfM_Doku38_Twitter_Online.pdf
Further information:
http://en.ejo.ch/tag/twitter-and-journalism-the-influence-of-the-social-web-on-
the-news