Real-time Viewer Feedback in the iTV Production Janez Zaletelj
Mladen Savić
Marko Meža
University of Ljubljana Trzaska 25, SI-1000 Ljubljana, Slovenia +386 14768 853
University of Ljubljana Trzaska 25, SI-1000 Ljubljana, Slovenia +386 14768 853
University of Ljubljana Trzaska 25, SI-1000 Ljubljana, Slovenia +386 14768 857
[email protected]
[email protected]
[email protected]
effectively use this new information. The collection of user feedback through the ipTV application will enable modeling and tracking of the TV viewers within the TV production, and personalized content production for the target viewer groups.
ABSTRACT This paper presents the concept of viewer feedback in the production of multi-channel, live TV shows. The IST-FP6 project “LIVE – Live Staging of Media Events” is developing a production support system which will provide personalized content recommendations as well as an overview about viewers’ feedback for the production team in real-time. The paper outlines a concept of a feedback system consisting of ipTV application for the consumer, the feedback collection and analysis components, and the voting control and feedback presentation modules in the TV production. The LIVE Feedback system enables the production team to observe preferences of the TV viewers in realtime, during the live production of the show, and on the other hand gives viewers the possibility to actively influence the creation of live TV content through voting.
User modeling and personalized content selection is used in a number of fields such as web documents, e-mail, news, multimedia, TV programs, jokes, music, etc [3]. Personalization within the interactive TV domain is usually done on the level of selecting a suitable programme or channel out of a set of available channels for the individual viewer, and this service is called personalized electronic programme guide (pEPG) [4]. To be able to provide recommendations, the system must derive user’s preferences towards content, which is done through analysis of viewer’s past interaction with the TV content. The recommender system must thus track what the user is watching, in other words it must collect viewers’ feedback. The content-based recommender then learns the profile of the user’s interest based on the features present in the past items that the user has explicitly rated or consumed. The content item is then compared to the user model and thus rated for each individual consumer.
Categories and Subject Descriptors H.5.3 [Group and Organization Interfaces]: Computersupported cooperative work H.3.4 [Systems and Software]: User profiles and alert services
The IST-FP6 project “LIVE – Live Staging of Media Events” [7] is developing a production support system which will provide personalized content recommendations as well as an overview about viewers’ feedback for the production team in real-time. The basic difference to the existing iTV recommender systems is that LIVE system provides personalization of the programme content itself during the live production [5], and not only selection amongst a set of available programme items. The basic requirement for live production of personalized TV content is that the channel TV audience is tracked in real-time. The LIVE system thus includes the Feedback collection and analysis services, which perform this task, and are outlined in this paper.
General Terms Algorithms, Design, Experimentation
Keywords TV production, Personalization, Interactive TV, Feedback
1. INTRODUCTION The introduction of interactive TV [1,2] opens up new opportunities in the existing world of TV services. The interactive TV applications enable consumers to actively provide feedback on the TV program. The consumers’ behavior and content and channel preferences can be tracked, collected by the application and fed-back to the content producer. The information on the TV audience preferences could be used for the planning of the TV program. However this information could be exploited also during the production of the live program such as sport transmission. This would require that the user feedback is collected in real time and that the whole content delivery process is adapted to
2. FEEDBACK SYSTEM FOR THE LIVE TV PRODUCTION While the today’s TV consumers are limited in their interaction with the TV to selecting different channels by switching, the interactive TV consumer is assisted by intelligent iTV application which provides new content and interactivity options. In order to provide personalized viewing experience, the iTV application might provide additional services:
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. EuroITV’09, June 3–5, 2009, Leuven, Belgium. Copyright 2009 ACM 978-1-60558-340-2/09/06...$5.00.
149
•
switching between sub-channels with different viewing angles or different coverage of the same event,
•
giving the viewer an option to express his feedback on the current program by voting or rating,
•
viewers, and to transfer the viewer actions (responses to the voting and channel switch information) to the TV production studio.
enabling the viewer to adjust his personal user profile which is used for personalized recommendations such as personalized electronic programme guides [4].
In this paper we concentrate on the real-time consumer feedback modalities, which were employed during live production as an important guide for content creation. Specifically two forms of consumers’ feedback were applied within LIVE system:
•
Feedback collection and analysis services were developed by the University of Ljubljana and provide real-time analysis about the current channel audience, their preferences, and responses to voting.
•
Explicit user feedback by voting. The viewer can choose one from the set of provided answers, which were defined on-the-fly in the live production.
•
Feedback application (GUI) was developed as a front-end application which allows setting up the analysis parameters and graphically displays the results to the production team.
•
Implicit user feedback by tracking channels. The switching between channels is collected by the system and the watching times are collected for each subchannel.
The consumer feedback is analyzed by the Feedback Analysis System and presented to the Director. The audience feedback plays a very important role in the new live production practices, which are being developed under a concept of “live staging of media events” [6]. While the current TV production involves lots of activities which are done off-line (in pre-production) and relies heavily on the pre-programmed schedule, the new concept of live staging on the other hand introduces flexibility in the content creation, which is based on real-time observation of happenings in the event, as well as realtime observation of the viewers’ acceptance of the content. The concept of live conducting requires a heavy support of automated systems for content preparation as well as systems to track and observe viewers’ feedback in real-time, which is the goal of the LIVE Feedback system.
Figure 1. Feedback System architecture. Users of the Feedback system play three main roles. These are related to the editor team, the director and the consumers. The director conducts the show, which is being watched by the consumers. The director is presented with information on feedback of consumers such as which channels he conducts are being watched as well as analysis of voting results. The editor’s team supports the director by actually initiating voting notifications, and setting up the view of the feedback analysis presentation.
The LIVE Feedback System enables the production team to observe in real-time behavior and preferences of the TV audience on the target channels. The real-time analysis of the migration of the Consumers between channels and their response to voting notifications provides an instant feedback on the acceptance of the LIVE channels by the audience, and allows the production team to react and possibly adjust the content of the TV show on-the-fly according to the audience feedback.
2.2 Preparation of the Feedback system Before the Olympic Games Field Trial, consumers were asked to fill out a questionnaire about their interest to the selected sport disciplines (32 disciplines). They were able to choose amongst either interested or not interested in a specific sport discipline. Additionally, they were asked a question on how much they are likely to watch a sport discipline events. They had to choose a level of interest on the discipline events from a 3 level scale (not likely to watch, maybe, very likely to watch). These answers were transformed into a 3-level interest profile of each of the Consumers.
The LIVE Feedback system allows for the first time in the TV production practice that the Director and the production team have full overview on their audience, and that they can even engage into bi-directional communication and interaction with the audience. Through the audience response (voting) on simple questions, the production team is able to get the feedback on the past or planned content of the show, which might prove as a very valuable tool for future live production practices.
2.1 Architecture of LIVE Feedback System Figure 1 shows a conceptual diagram of a LIVE Feedback system setup, which was employed during the production of the Olympic Games 2008 test live transmission at the Austrian Broadcasting Company (ORF). The main components are the following: •
•
Viewers were assigned to user groups based on their expressed interests and the demographic data, which resulted in a total of 5 demographic groups (based on age and gender) and 32 interest groups (based on sport disciplines).
Consumer’s ipTV application which is capable of displaying the notifications (questions) sent by the Director to the specific segment of the audience, and to return feedback information.
3. ANALYSIS OF VIEWERS’ FEEDBACK Real-time analysis of the channel audience was performed during the trial.
Real-time notification and feedback channel which transfers the viewers’ actions to the TV production office. The ipTV solution of Telekom Austria was modified and integrated with the LIVE production support system to display notifications (messages) from the production to the
3.1 Channel statistics The system provides an overview of number of users on each of the channels. Let denote the number of users on a i-th channel at a time t. Let , 0,1 is a binary function which
150
defines if a k-th user is watching i-th channel at time t. Number of users per channel is calculated for each of the 5 interlinked channels for which the switching data was available.
channel viewers, how they rate his show, the results of voting and channel statistics (Figure 2).
The dynamic aspect of the channel audience was also calculated as a change rate of viewers $ % ⁄. A Gaussian kernel was used for smoothing the data.
3.2 User Group analysis
Similarly, the number of viewers was counted for each of the 32 user groups, which gave an overview on the preferences of the current audience towards different sport disciplines. The group profile was derived from the individual profiles of the viewers which belonged to a certain group. The consumer profile is represented by a vector , , where , 0,0.5,1 gives the preference of the user k to the attribute (sport discipline) l. The , 0,1 represents a relevance of the attribute l to the user k.
The group profile is calculated so that the average preference for the attribute l over all consumers who have provided preference value is calculated first as
Figure 3. Feedback Application window with histogram and time-based comparison of target channels.
, · , !" ,
The Feedback Application window consists of two panes. The left pane contains a tree-structure for setting the graph parameters and right pane contains the graphs created using the settings from the left pane. Different feedback presentation modes and user options are described next.
Average relevance of a sport discipline to the user group is calculated as a relative number of users who have rated this sport discipline and denoted as . Finally, relative importance of the attribute (sport discipline) to the whole user group is calculated as the product · . The final vector which represents a group profile in terms of attribute preferences is # · .
•
An example of the weighted group profile # for the audience group “Men” is shown in Figure 2, indicating that the football is the most popular sport between men audience.
•
Men group 0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
0,9
1
Football Athletics Swimming Tennis Cycling Boxing Table Tennis Basketball Sailing Volleyball Triathlon Handball Gymnastics Archery Water Jump Modern Pentathlon Shooting Watersports Baseball Riding Judo Badminton Synchronized Swimming Rowing Weightlifting Taekwondo Water Polo Fencing Canoe Field Hockey Wrestling Softball
•
Voting and rating (explicit feedback). The viewers are able to express their opinion whether they like the current program or not (rating) and are able to answer to various voting polls. The results are shown in a graphical format as a rating chart (histogram or pie chart), and voting result char (histogram or pie chart). Channel comparison graph based on implicit consumer feedback. The watching times of Consumers are collected for each channel and time-based channel viewing statistics is computed. This feedback is even more powerful in combination with information on consumer profiles. There are two types of graphs available for displaying channel viewing statistics: time-based graph and histogram (Figure 3). Observation of the target groups. The Director is able to track the distribution of several pre-defined user groups on the target channel, which provides information about demographic or interest distribution of viewers on a target channel.
Time-based chart provides some additional options and functionalities. To the right of the graph title there is a set of buttons, which are used for displaying graph markers. Markers denote a moment in time when something significant happened (the director initiated a voting, sent a message etc.). There are three types of markers: for votings, for messages and for timeintervals. The markers allow the team to easily observe the effect of some action (such as initiation of voting) on the audience over time.
Figure 2. Audience profile for “Men”.
4. PRESENTATION OF FEEDBACK IN THE CONTROL ROOM Feedback Application is an end-user application for professional users, which analyses and displays various modalities of feedback from the entire TV audience. Feedback Application was designed in such a way that the Video conductor is able to set up feedback analysis parameters and to review the real-time aggregated statistical data on his audience, including preferences of the
151
viewer is able to actively influence the content of the live TV program such as transmission of the sport event. This new concept of interactive TV production was tested within EU project »LIVE« during the Olympic Games 2008. The evaluation of the concept included 480 households in Austria which were able to receive multi-channel live coverage of the Olympic Games, produced by ORF and supported by the presented system. The results of the evaluation shown the great promise of this concept, and the acceptance of the presented tools by the professional users in the TV production. The LIVE Feedback System is an important advancement on the road to fully interactive and personalized TV. On one hand it brings into the production office the new real-time information on audience behavior and preferences, which can also be employed for future content planning as well as for analysis of acceptance and reach of particular content. On the other hand, the response from the Consumers who were watching LIVE channels shows that they really appreciate new interactivity options and that they feel, that through the voting feedback they can really influence the content of the show and in this way they became more engaged with the TV programme.
Figure 4: Voting and message markers in viewing statistics time graph
4.1 Dynamic behaviour of the Audience Histogram charts display trend markers to make it easier for the user to have a feeling about whether the number of viewers is increasing or decreasing. The Feedback GUI remembers the past values of viewers on different channels and compares them to the present values. It then draws line indicators where previous values used to be. The lines are coloured red and green, depending whether the number of consumers decreased or increased.
6. ACKNOWLEDGMENTS The LIVE project is a joint research and development effort funded by the European Commission within the 6th framework of the IST under grant number FP6-27312. All statements in this work reflect the personal ideas and opinions of the authors and not necessarily the opinions of the European Commission.
Besides the markers, the time-graph provides an option of alarming the user about some important trends (positive or negative) in the audience. An example of a negative trend would be a sudden drop in number of viewers on a channel or no viewers on a channel. When an event occurs, the system alarms the user by painting a colored frame around the chart on which the event occurred. The frame is colored red for negative events and green for positive.
REFERENCES [1] A. Hartmann, Producing Interactive Television, Charles River Media, Inc., Hingham, MA, 2002. [2] ITV Production Standards Committee, iTV Standards Links, http://itvstandards.org/iTVPublic/standards.aspx [3] A. Kobsa, “User Modeling and user adapted interaction”, User Modeling and user adapted interaction, Vol. 11, p.4963, 2001 [4] D. O’Sullivan, B. Smyth, D.C. Wilson, K. McDonald, and A. Smeaton, “Improving the quality of the personalized electronic program guide”, User Modeling and user adapted interaction, Vol. 14, p.5-36, 2004 [5] Janez Zaletelj, “Recommender System for the Multi-channel TV Production”, in M.Tscheligi, M.Obrist, and A.Lugmayr(Eds.): EuroITV 2008, LNCS, Vol. 5066, pp. 102-106, 2008
Figure 5: A typical histogram with trend markers. Green markers denote the increase in the number of viewers, wile, red markers denote decrease.
[6] Stefan M. Grünvogel, Richard Wages, Tobias Bürger and Janez Zaletelj, “A Novel System for Interactive Live TV”, in L.Ma, R.Nakatsu, and M.Rauterberg (Eds.): ICEC 2007, LNCS 4740, pp. 193-204, 2007 [7] LIVE – Live staging of media events; project web-site: http://www.ist-live.org – Last visited: 1.03.2009
5. CONCLUSIONS The ipTV technology enables new interaction possibilities between the TV production and the TV consumer. Through the real-time feedback channel and analysis of feedback, the TV
152
