Building public service recommenders: Logbook of a journey
RecSys 2021 Building public service recommenders: Logbook of a journey
Authors: Christina Boididou, BBC | Di Sheng, BBC | Felix J Mercer Moss, Datalab BBC | Alessandro Piscopo, BBC
Abstract: The BBC is one of the world's leading broadcasters, regularly producing large volumes of content including video, audio and text, spanning topics such as news, sport, education, and entertainment. In order to fulfil its public service remit, the BBC must cater to the full breadth of interests within its audiences, providing each member with the most relevant and engaging content for them.
Up until now, manual curation has been the principal approach used by the organisation to achieve this end. However, this strategy does not scale and is ineffective at tailoring personalised content experiences.
Modern recommender systems appear to represent relevant and appropriate solutions to the problem of scalable algorithmic curation. How can such systems be developed to in a public service organisation?
The core characteristics of public service can be difficult to define, but descriptions invariably include ideas such as independence, impartiality, diversity and inclusiveness.
To support the BBC in developing and deploying responsible recommendations at scale our team, Datalab, has taken an approach centred around the collaboration across teams and professional roles (product, editorial, and data), and around the concept of public service AI.
In this talk, we present this approach and reflect upon the lessons learnt in the years since the creation of our team.
DOI: https://doi.org/10.1145/3460231.3474614
Видео Building public service recommenders: Logbook of a journey канала ACM RecSys
Authors: Christina Boididou, BBC | Di Sheng, BBC | Felix J Mercer Moss, Datalab BBC | Alessandro Piscopo, BBC
Abstract: The BBC is one of the world's leading broadcasters, regularly producing large volumes of content including video, audio and text, spanning topics such as news, sport, education, and entertainment. In order to fulfil its public service remit, the BBC must cater to the full breadth of interests within its audiences, providing each member with the most relevant and engaging content for them.
Up until now, manual curation has been the principal approach used by the organisation to achieve this end. However, this strategy does not scale and is ineffective at tailoring personalised content experiences.
Modern recommender systems appear to represent relevant and appropriate solutions to the problem of scalable algorithmic curation. How can such systems be developed to in a public service organisation?
The core characteristics of public service can be difficult to define, but descriptions invariably include ideas such as independence, impartiality, diversity and inclusiveness.
To support the BBC in developing and deploying responsible recommendations at scale our team, Datalab, has taken an approach centred around the collaboration across teams and professional roles (product, editorial, and data), and around the concept of public service AI.
In this talk, we present this approach and reflect upon the lessons learnt in the years since the creation of our team.
DOI: https://doi.org/10.1145/3460231.3474614
Видео Building public service recommenders: Logbook of a journey канала ACM RecSys
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
Tops, Bottoms, and Shoes: Building Capsule Wardrobes via Cross-Attention Tensor NetworkPS 5: Latent Factor Models and Aggregation Operators for Collaborative Filtering in Reciprocal RecomPS2: Translation-based factorization machines for sequentialWorkshop on Context-Aware Recommender SystemsPaper Session 4: Domain Adaptation in Display Advertising: An Application for Partner Cold-StartMitigating Confounding Bias in Recommendation via Information BottleneckPS 7: Eliciting pairwise preferences in recommender systems Saikishore KallooriPS 6: Judging similarity: a user-centric study of related item recommendations Yuan YaoRecSys 2015 Session 4b: AlgorithmsBoosting Local Recommendations With Partially Trained Global ModelPessimistic Reward Models for Off-Policy Learning in RecommendationLearning a voice-based conversational recommender using offline policy optimizationSession 9: Timely Personalization at Peloton:System and Algorithm for Boosting Time Relevant ContentRecSys 2020 Session P2B: Evaluating and Explaining RecommendationsRecSys 2020 Session P5B: Real World Applications IIPrivacy Preserving Collaborative Filtering by Distributed MediationRecSys 2016: Paper Session 7 - Behaviorism is Not Enough: Better RecommendationsRecSys 2016: Paper Session 5 - Mechanism Design for Personalized Recommender SystemsRecSys 2020 Session P5A: Real World Applications IIDebiased Off-Policy Evaluation for Recommender Systems