Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Learning and Optimization of Implicit Negative Feedback for Industrial Short-video Recommender System

Yunzhu Pan, Nian Li, Chen Gao, Jianxin Chang, Yanan Niu, Yang Song, Depeng Jin, Yong Li

TL;DR

This paper deploys a feedback-aware encoding module that extracts user preferences, taking the impact of context into consideration, and designs a multi-objective prediction module which well distinguishes the relation and differences among different model objectives in the short-video recommendation.

Abstract

Short-video recommendation is one of the most important recommendation applications in today's industrial information systems. Compared with other recommendation tasks, the enormous amount of feedback is the most typical characteristic. Specifically, in short-video recommendation, the easiest-to-collect user feedback is the skipping behavior, which leads to two critical challenges for the recommendation model. First, the skipping behavior reflects implicit user preferences, and thus, it is challenging for interest extraction. Second, this kind of special feedback involves multiple objectives, such as total watching time and skipping rate, which is also very challenging. In this paper, we present our industrial solution in Kuaishou, which serves billion-level users every day. Specifically, we deploy a feedback-aware encoding module that extracts user preferences, taking the impact of context into consideration. We further design a multi-objective prediction module which well distinguishes the relation and differences among different model objectives in the short-video recommendation. We conduct extensive online A/B tests, along with detailed and careful analysis, which verify the effectiveness of our solution.

Learning and Optimization of Implicit Negative Feedback for Industrial Short-video Recommender System

TL;DR

This paper deploys a feedback-aware encoding module that extracts user preferences, taking the impact of context into consideration, and designs a multi-objective prediction module which well distinguishes the relation and differences among different model objectives in the short-video recommendation.

Abstract

Short-video recommendation is one of the most important recommendation applications in today's industrial information systems. Compared with other recommendation tasks, the enormous amount of feedback is the most typical characteristic. Specifically, in short-video recommendation, the easiest-to-collect user feedback is the skipping behavior, which leads to two critical challenges for the recommendation model. First, the skipping behavior reflects implicit user preferences, and thus, it is challenging for interest extraction. Second, this kind of special feedback involves multiple objectives, such as total watching time and skipping rate, which is also very challenging. In this paper, we present our industrial solution in Kuaishou, which serves billion-level users every day. Specifically, we deploy a feedback-aware encoding module that extracts user preferences, taking the impact of context into consideration. We further design a multi-objective prediction module which well distinguishes the relation and differences among different model objectives in the short-video recommendation. We conduct extensive online A/B tests, along with detailed and careful analysis, which verify the effectiveness of our solution.
Paper Structure (27 sections, 9 equations, 4 figures, 4 tables)

This paper contains 27 sections, 9 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. The proposed system
  3. Feedback-aware Sequential Encoder
  4. Embedding Layer
  5. Self-Attention Encoder
  6. Context Feature Embedding Layer
  7. Feature construction.
  8. Feature transformation.
  9. Feature importance alignment.
  10. Multi-feedback Prediction Layer
  11. Training and Serving
  12. Joint training
  13. Fusion strategy for online serving
  14. Online Experiments
  15. Overall performance of A/B test
  16. ...and 12 more sections

Figures (4)

  • Figure 1: Illustration of our proposed system.
  • Figure 2: The performance improvement trend of our model in a one-week window. (a) The improvement trend of user's usage duration. (b) The improvement trend of video's playing duration on single-column pages. (c) The improvement trend of the main page's players and double-column page's visitors. (d) The improvement trend of the number of daily active users. (e) The improvement trend of the reduction number of similar recommendations, of which the lower number the better performance.
  • Figure 3: Daily active users (DAU) and user's average using time (Avg Use time) on different user engagement levels.
  • Figure 4: Relationship between negative feedback and diversity entropy in the long-term aspect.