Watch-It-Next: A Contextual TV Recommendation System
As consumers of television are presented with a plethora of available programming, improving recommender systems in this domain is becoming increasingly important. Television sets, though, are often shared by multiple users whose tastes may greatly vary.
- PDF / 1,006,969 Bytes
- 16 Pages / 439.37 x 666.142 pts Page_size
- 39 Downloads / 183 Views
Yahoo Labs, Haifa, Israel {michala,eshcar,akagian,raz}@yahoo-inc.com 2 Outbrain, Netanya, Israel [email protected] 3 Pontis, Ra’anana, Israel [email protected]
Abstract. As consumers of television are presented with a plethora of available programming, improving recommender systems in this domain is becoming increasingly important. Television sets, though, are often shared by multiple users whose tastes may greatly vary. Recommendation systems are challenged by this setting, since viewing data is typically collected and modeled per device, aggregating over its users and obscuring their individual tastes. This paper tackles the challenge of TV recommendation, specifically aiming to provide recommendations for the next program to watch following the currently watched program the device. We present an empirical evaluation of several recommendation methods over large-scale, real-life TV viewership data. Our extentions of common state-of-theart recommendation methods, exploiting the current watching context, demonstrate a significant improvement in recommendation quality.
1
Introduction
In recent years, online experiences have made increasing use of recommendation technology and user modeling techniques. Media sites recommend what to read, music streaming sites recommend what to listen to, VOD subscription services recommend what to watch, and more. In many of these instances, recommendation effectively transforms an online service to be personalized - tailored to its single or primary user. This occurs when recommendations are exposed in an experience consumed through a personal device such as a smartphone or a laptop, or when they are exposed on personal accounts such as a social network account. However, in other cases recommendation technology is applied in settings where multiple users share an application or device. Examples include game consoles or high end smart TVs in household living rooms, family accounts in VOD subscription services or eCommerce sites, and shared desktops or tablets in homes. These multi-user cases represent a challenge to recommender systems, as recommendations given to one user may actually be more suitable for another c Springer International Publishing Switzerland 2015 A. Bifet et al. (Eds.): ECML PKDD 2015, Part III, LNAI 9286, pp. 180–195, 2015. DOI: 10.1007/978-3-s319-23461-8 12
Watch-It-Next: A Contextual TV Recommendation System
181
user. In general, the utility of recommendations in multi-user settings is lower, sometimes to the point of frustrating users when the recommendations they receive are dominated by items suitable for others. This paper tackles the TV recommendation problem by applying context to implicitly disambiguate the user, or users, that are watching it. Specifically, we address the household smart TV situation, where a television set runs software that identifies what is watched on it, and taps that knowledge to recommend to its owners what they should watch at any given time. Obviously, individual household members may have very different viewing habit
Data Loading...