First study on a complete dataset of Tweet
Speech presented during the 4th edition of Transforming Audiences Conference, University of Westminister - 3 Septermber 2013
Generative Artificial Intelligence: How generative AI works.pdf
Second Screen and Political Talk-Shows: Measuring and Understanding the Italian Participatory «Couch Potato»
1. Fabio [.] Giglietto [@uniurb.it]
Department of Communication Studies and Humanities | Università di Urbino Carlo Bo
SECOND SCREEN AND POLITICAL TALK-SHOWS:
MEASURING AND UNDERSTANDING THE ITALIAN
PARTICIPATORY «COUCH POTATO»
SOCIAL MEDIA - THE FOURTH ANNUAL TRANSFORMING AUDIENCES CONFERENCE
WESTIMIINISTER UNIVERSITY, LONDON, 2-3 SEPTEMBER 2013
2. Outline
• Dataset
• 3 research questions
– Methodology
– Data Analysis
– Results
• Conclusions
3. Dataset
• From 30th of August 2012 to 30th June 2013
• 11 political talk-shows
• Hashtags: #ballarò or #ballaro, #portaaporta,
#agorarai, #ultimaparola, #serviziopubblico,
#inmezzora, #infedele or #linfedele, #ottoemezzo,
#omnibus, #inonda, #piazzapulita
• Complete dataset from Twitter firehose
(DiscoverText + GNIP)
• Raw n. of Tweets collected: 2,489,669
4. Research Questions
• RQ1: is the activity on Twitter statistically
correlated with TV audience?
• RQ2: what kind of TV scene corresponds to
the highest engagement on Twitter?
• RQ3: Twitter commentaries on political talk-
shows are at the cross-road between political
and audience participation. Which of these
two types of participation is the most
common and how it is played?
6. RQ1 dataset
• Subset of Tweets (1) created during the on air
time of the episodes (+15 mins) and (2)
containing the corresponding program
#hashtag (n= 1,881,873)
• 1,077 aired episodes with respective average
audience and rating as estimated by Auditel
• Twitter metrics for each episode (Tweets,
contributors, reach, ReTweet, Reply, Tweet-
per-minute, contributors-per-minute)
12. Results (1/2)
• Over the eight different metrics tested, the
observed correlation coefficient with the
audience was > 0.5
• The rate of Tweet per minute (TPM) and
contributors per minute (CPM) correlate
remarkably well (when log transformed
respectively r=0.83 and 0.86 with p < .01)
13. Results (2/2)
• A multiple regression model based on the (1)
average audience of previously aired episodes,
(2) CPM and (3) networked publics variable*,
explained 96% of the variance in the audience
• Taking all other variables constant, we expect
an increase of 0.37% in audience for an
increase of 1% in average CPM
* representing the inclination of the audience base of a show to contribute to the
conversation with the official hashtag while the show is on air
14. WHAT KIND OF TV SCENE
CORRESPONDS TO THE HIGHEST
ENGAGEMENT ON TWITTER?
RQ2
15. Definitions
• Period <- span of n days between main
political events of the season
• Original Tweets < Tweet-(RT+Reply)
• Engagement < Peaks in Original Tweets
• Window < span of n minutes around the peak
• TV scene < excerpt of a TV program aired
during a window
16. Dataset
• Twitter metrics (tweet, rt, reply, contributors,
reach, original tweets) by minutes from 30
August 2012 to 30 June 2013 (n=439,204)
20. Methods
• Peaks detection (Marcus et al 2011)
• Textmining of Tweets created during each
windows to find the top 5 frequent term (td-idf)
extraction and automatic labeling of the peaks
• Triangluated content analysis of a random sample
of 22 peaks (2 for period): (1) Tweets created
during the peak, (2) peak’s label and (3)
correponding TV events (retived via web
streaming)
22. Peaks detected in the outlier episode
POLVERINI, PIAZZAPULITA, GOCCE, LAZIO, RENATA POLVERINI, PIAZZAPULITA, FOTOGRAFO, LAZIO, GIUNTA
full list of peaks with automatic labels
23. Preliminary results (1/2)
• During the whole season: more than half of the
analyzed peaks correspond to scene with
controversial guests, screams, fights, squabbles
and occasional bad words
• During the first part of the season: scandals, anti-
politics and the critics agaist the austerity
measures decided by Monti’s government
• During the campaign: Silvio Berlusconi (whenever
present as a guest, imitated by a comedian or
critizied by his opponents)
24. Preliminary results (2/2)
• After the elections and before the formation
of the new government: any opportunity for
critizing the democratic party and call for a
more direct participation
25. WHICH TYPES OF PARTICIPATION
(AUDIENCE/POLITICAL ) IS THE MOST FREQUENT?
HOW IT IS PLAYED?
RQ3
26. Methods
• Triangulated content analysis of a random
sample of 22 peaks (2 for period): (1) Tweets
created during the window, (2) peak’s label
and (3) correponding TV events (retived via
web streaming)
• Dataset of all Tweets in each windows
(n=7,394) excluding ReTweets
• Codeset > modified version of a code matrix
originally developed by Wohn & Na, 2011
28. Preliminary results (1/2)
• Audience participation prevails during scene
with controversial guests, screams, fights,
squabbles and bad words
• Political participation prevails during
interviews
• Both for audience and political participation,
information often express opinions by
hilighting agreed, disagreed or underline
wrong sentences
29. Preliminary results (2/2)
• Opinions are often addressed to a general
imagined audience but sometimes are directly
addressed (even using the @) to a guest or to
the host of the show as in an imaginary peer
to peer conversation
• Emotions are often hilighted by the use of
multple exclamation marks
30. Limitation and future works
• RQ1 results are applicable to the analyzed
genre only (movies, tv series, etc probably
works in a different way)
• In order to better address RQ2 all peaks
should be analyzed with the same method
• RQ3 should be tested not only around peaks
• In both cases it would be good to have more
than one coder in order to get intercoder
reliability
31. Thanks for the attention!
Working paper on RQ1 is available on SSRN
Please address comments, suggestions and
remarks to @fabiogiglietto
Notas do Editor
What I will present today is, at the best of knowledge, the first study on an entire TV season of Tweets related to a genre (political talk shows)
But, at the same time, TV has never been as social as it is today