Foresight Prediction Enhanced Live-Streaming Recommendation
Jiangxia Cao, Ruochen Yang, Xiang Chen, Changxin Lao, Yueyang Liu, Yusheng Huang, Yuanhao Tian, Xiangyu Wu, Shuang Yang, Zhaojie Liu, Guorui Zhou
TL;DR
<3-5 sentence high-level summary> The paper tackles real-time live-streaming recommendation by identifying highlight moments and addressing the challenge of unseen future content. It introduces Segment Semantic Prediction, which quantizes segment-level multimodal embeddings into Semantic IDs (Sids) using a K-Means codebook, and then forecasts the next Sid with a transformer-based encoder–decoder. These foresight signals are incorporated into the ranking stage as enhanced features, with streaming training to handle continuous content. Extensive offline and online experiments show large gains in AUC/GAUC and engagement metrics, demonstrating practical benefits for user experience and revenue on large-scale platforms.
Abstract
Live-streaming, as an emerging media enabling real-time interaction between authors and users, has attracted significant attention. Unlike the stable playback time of traditional TV live or the fixed content of short video, live-streaming, due to the dynamics of content and time, poses higher requirements for the recommendation algorithm of the platform - understanding the ever-changing content in real time and push it to users at the appropriate moment. Through analysis, we find that users have a better experience and express more positive behaviors during highlight moments of the live-streaming. Furthermore, since the model lacks access to future content during recommendation, yet user engagement depends on how well subsequent content aligns with their interests, an intuitive solution is to predict future live-streaming content. Therefore, we perform semantic quantization on live-streaming segments to obtain Semantic ids (Sid), encode the historical Sid sequence to capture the author's characteristics, and model Sid evolution trend to enable foresight prediction of future content. This foresight enhances the ranking model through refined features. Extensive offline and online experiments demonstrate the effectiveness of our method.