VideoRecSys 2023: First Workshop on Large-Scale Video Recommender Systems Khushhall Chandra Mahajan Meta Inc., USA khushhall@meta.com Amey Porobo Dharwadker Meta Inc., USA ameydhar@meta.com Saurabh Gupta Meta Inc., USA saurabg@meta.com Brad Schumitsch Meta Inc., USA bschumitsch@meta.com ABSTRACT The demand for personalized video recommendations has grown exponentially with the widespread use of video content across vari- ous domains, including entertainment, e-commerce, education and social media. The explosive growth of video content on the internet, combined with the ubiquitous availability of high-speed internet and advancements in mobile camera technology have made it easier than ever for users to create, access and consume videos. With the proliferation of online social media applications like Instagram, YouTube, Facebook and TikTok, the need for large-scale video rec- ommendation systems which can provide users with personalized and relevant recommendations has increased. However, building effective and scalable video recommender systems for large-scale applications is a challenging task due to several technical and opera- tional factors such as the huge volume of video content, the diversity of user preferences, noise and biases in underlying data and the need for real-time recommendations. At the same time, building a fair recommendation system which can distribute content from niche interests and emerging creators besides more popular content is very important to balance the demand and supply side of the ecosystem. This makes large-scale video recommendations an in- teresting, rapidly evolving and challenging problem which requires more discussions around various nuanced topics. The goal of this workshop is to provide a platform for researchers, practitioners, and industry experts to discuss and share insights on the latest trends, challenges and opportunities in the field of large-scale video recommendations, where the corpus of content to recommend from is in the order of hundreds of millions or even tens of billions. We hope that this workshop will foster collaborations in the growing recommender systems community and spark new ideas for future research in this exciting and rapidly evolving field. CCS CONCEPTS • Information systems → Recommender systems; Personal- ization. Permission to make digital or hard copies of part or all 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. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner/author(s). RecSys ’23, September 18–22, 2023, Singapore, Singapore © 2023 Copyright held by the owner/author(s). ACM ISBN 979-8-4007-0241-9/23/09. https://doi.org/10.1145/3604915.3608762 KEYWORDS Recommender system, Video recommendation ACM Reference Format: Khushhall Chandra Mahajan, Amey Porobo Dharwadker, Saurabh Gupta, and Brad Schumitsch. 2023. VideoRecSys 2023: First Workshop on Large- Scale Video Recommender Systems. In Seventeenth ACM Conference on Rec- ommender Systems (RecSys ’23), September 18–22, 2023, Singapore, Singapore. ACM, New York, NY, USA, 3 pages. https://doi.org/10.1145/3604915.3608762 1 INTRODUCTION Video recommender systems have become an essential part of our daily lives. With the increasing popularity of online platforms, such as Instagram, YouTube, Netflix, Facebook, Amazon and Tik- Tok, users are overwhelmed with a vast amount of video content to choose from. Recommender systems address this challenge by suggesting personalized video recommendations based on users’ preferences and interests. The goal of video recommender systems is to predict which videos a user is likely to watch and enjoy, given their past viewing history and interactions with the platform. This is typically achieved by analyzing the user’s behavior data, such as clicks, likes, and watch time, and using this information to build a model that can accurately predict their future preferences. Despite the significant progress made in the field of video rec- ommender systems, there are still several challenges and problem areas that remain relevant to the recommender systems community. One of the main challenges is the cold start problem [6, 18], where new users or items have limited or no interaction data available. To address this issue, various methods have been proposed, including leveraging user demographic data and content metadata to infer preferences and content. Collaborative filtering techniques such as matrix factorization and graph neural network models have also been explored to generate accurate recommendations for new users and videos [3, 22]. Another challenge is the issue of promoting diversity in recom- mendations. Recommender systems often tend to suggest popu- lar videos or videos that are similar to what the user has already watched [7], leading to a lack of diversity in the recommendations. This in turn, can lead to a user experiencing boredom or frustra- tion with the platform and negatively impact their engagement. To address this challenge, researchers have focused on improving the exploration-exploitation trade-off [2, 5, 17, 20], which can bet- ter attend to users’ long-term interests [13, 16] and improve their overall experience with the recommendation system. Strategies to 1269