Personalized recommendation of linear content on interactive TV platforms

Beating the cold start and noisy implicit user feedback

David Zibriczky, Balázs Hidasi, Zoltán Petres, D. Tikk

Research output: Chapter in Book/Report/Conference proceedingConference contribution

6 Citations (Scopus)

Abstract

Recommender systems in TV applications mostly focus on the recommendation of video-on-demand (VOD) content, although the major part of users' content consumption is realized on linear channel programs (live or recorded), termed EPG programs. The accurate collaborative filtering algorithms suitable for VOD recommendation cannot be directly carried over for EPG program recommendation. First, EPG program recommendation features the cold start problem; a significant part of EPG programs are new in the system. Second, and more importantly, without explicit user feedbacks (ratings) the algorithms have to model user preference based on the noisy and less directly interpretable implicit user feedbacks. In this paper, we present several approaches that overcome these difficulties, by applying pre-filtering on noisy low-level data and taking into account channel preferences of users and program metadata if available to cope with the cold start. Using time-dependent tensor factorization approaches, the temporal preferences of users are also reflected in recommendation, that also hints on the person watching the TV. Experiments were performed on a dataset of SaskTel, a Canadian IPTV service provider using Microsoft Mediaroom middleware.

Original languageEnglish
Title of host publicationCEUR Workshop Proceedings
Volume872
Publication statusPublished - 2012
Event20th Conference on User Modeling, Adaptation, and Personalization, UMAP 2012 - Montreal, QC, Canada
Duration: Jul 16 2012Jul 20 2012

Other

Other20th Conference on User Modeling, Adaptation, and Personalization, UMAP 2012
CountryCanada
CityMontreal, QC
Period7/16/127/20/12

Fingerprint

Video on demand
Feedback
IPTV
Collaborative filtering
Recommender systems
Metadata
Middleware
Factorization
Tensors
Experiments

Keywords

  • Alternating least squares
  • Cold start problem
  • Implicit feedback collaborative filtering
  • Matrix factorization
  • Tensor factorization
  • Time-dependent modeling

ASJC Scopus subject areas

  • Computer Science(all)

Cite this

Personalized recommendation of linear content on interactive TV platforms : Beating the cold start and noisy implicit user feedback. / Zibriczky, David; Hidasi, Balázs; Petres, Zoltán; Tikk, D.

CEUR Workshop Proceedings. Vol. 872 2012.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Zibriczky, D, Hidasi, B, Petres, Z & Tikk, D 2012, Personalized recommendation of linear content on interactive TV platforms: Beating the cold start and noisy implicit user feedback. in CEUR Workshop Proceedings. vol. 872, 20th Conference on User Modeling, Adaptation, and Personalization, UMAP 2012, Montreal, QC, Canada, 7/16/12.
Zibriczky, David ; Hidasi, Balázs ; Petres, Zoltán ; Tikk, D. / Personalized recommendation of linear content on interactive TV platforms : Beating the cold start and noisy implicit user feedback. CEUR Workshop Proceedings. Vol. 872 2012.
@inproceedings{15ea683d36fc4d67847f51ff17481213,
title = "Personalized recommendation of linear content on interactive TV platforms: Beating the cold start and noisy implicit user feedback",
abstract = "Recommender systems in TV applications mostly focus on the recommendation of video-on-demand (VOD) content, although the major part of users' content consumption is realized on linear channel programs (live or recorded), termed EPG programs. The accurate collaborative filtering algorithms suitable for VOD recommendation cannot be directly carried over for EPG program recommendation. First, EPG program recommendation features the cold start problem; a significant part of EPG programs are new in the system. Second, and more importantly, without explicit user feedbacks (ratings) the algorithms have to model user preference based on the noisy and less directly interpretable implicit user feedbacks. In this paper, we present several approaches that overcome these difficulties, by applying pre-filtering on noisy low-level data and taking into account channel preferences of users and program metadata if available to cope with the cold start. Using time-dependent tensor factorization approaches, the temporal preferences of users are also reflected in recommendation, that also hints on the person watching the TV. Experiments were performed on a dataset of SaskTel, a Canadian IPTV service provider using Microsoft Mediaroom middleware.",
keywords = "Alternating least squares, Cold start problem, Implicit feedback collaborative filtering, Matrix factorization, Tensor factorization, Time-dependent modeling",
author = "David Zibriczky and Bal{\'a}zs Hidasi and Zolt{\'a}n Petres and D. Tikk",
year = "2012",
language = "English",
volume = "872",
booktitle = "CEUR Workshop Proceedings",

}

TY - GEN

T1 - Personalized recommendation of linear content on interactive TV platforms

T2 - Beating the cold start and noisy implicit user feedback

AU - Zibriczky, David

AU - Hidasi, Balázs

AU - Petres, Zoltán

AU - Tikk, D.

PY - 2012

Y1 - 2012

N2 - Recommender systems in TV applications mostly focus on the recommendation of video-on-demand (VOD) content, although the major part of users' content consumption is realized on linear channel programs (live or recorded), termed EPG programs. The accurate collaborative filtering algorithms suitable for VOD recommendation cannot be directly carried over for EPG program recommendation. First, EPG program recommendation features the cold start problem; a significant part of EPG programs are new in the system. Second, and more importantly, without explicit user feedbacks (ratings) the algorithms have to model user preference based on the noisy and less directly interpretable implicit user feedbacks. In this paper, we present several approaches that overcome these difficulties, by applying pre-filtering on noisy low-level data and taking into account channel preferences of users and program metadata if available to cope with the cold start. Using time-dependent tensor factorization approaches, the temporal preferences of users are also reflected in recommendation, that also hints on the person watching the TV. Experiments were performed on a dataset of SaskTel, a Canadian IPTV service provider using Microsoft Mediaroom middleware.

AB - Recommender systems in TV applications mostly focus on the recommendation of video-on-demand (VOD) content, although the major part of users' content consumption is realized on linear channel programs (live or recorded), termed EPG programs. The accurate collaborative filtering algorithms suitable for VOD recommendation cannot be directly carried over for EPG program recommendation. First, EPG program recommendation features the cold start problem; a significant part of EPG programs are new in the system. Second, and more importantly, without explicit user feedbacks (ratings) the algorithms have to model user preference based on the noisy and less directly interpretable implicit user feedbacks. In this paper, we present several approaches that overcome these difficulties, by applying pre-filtering on noisy low-level data and taking into account channel preferences of users and program metadata if available to cope with the cold start. Using time-dependent tensor factorization approaches, the temporal preferences of users are also reflected in recommendation, that also hints on the person watching the TV. Experiments were performed on a dataset of SaskTel, a Canadian IPTV service provider using Microsoft Mediaroom middleware.

KW - Alternating least squares

KW - Cold start problem

KW - Implicit feedback collaborative filtering

KW - Matrix factorization

KW - Tensor factorization

KW - Time-dependent modeling

UR - http://www.scopus.com/inward/record.url?scp=84891794086&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84891794086&partnerID=8YFLogxK

M3 - Conference contribution

VL - 872

BT - CEUR Workshop Proceedings

ER -