Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Trinity: A Scenario-Aware Recommendation Framework for Large-Scale Cold-Start Users

Wenhao Zheng, Wang Lu, Fangshuang Tang, Yiyang Lu, Jun Yang, Pengcheng Xiong, Yulan Yan

TL;DR

This paper proposes Trinity, a framework embodying the synergistic integration of feature engineering, model architecture, and stable model updating that achieves substantial improvements in addressing the combined challenge of new users in new scenarios.

Abstract

Early-stage users in a new scenario intensify cold-start challenges, yet prior works often address only parts of the problem through model architecture. Launching a new user experience to replace an established product involves sparse behavioral signals, low-engagement cohorts, and unstable model performance. We argue that effective recommendations require the synergistic integration of feature engineering, model architecture, and stable model updating. We propose Trinity, a framework embodying this principle. Trinity extracts valuable information from existing scenarios while ensuring predictive effectiveness and accuracy in the new scenario. In this paper, we showcase Trinity applied to a billion-user Microsoft product transition. Both offline and online experiments demonstrate that our framework achieves substantial improvements in addressing the combined challenge of new users in new scenarios.

Trinity: A Scenario-Aware Recommendation Framework for Large-Scale Cold-Start Users

TL;DR

This paper proposes Trinity, a framework embodying the synergistic integration of feature engineering, model architecture, and stable model updating that achieves substantial improvements in addressing the combined challenge of new users in new scenarios.

Abstract

Early-stage users in a new scenario intensify cold-start challenges, yet prior works often address only parts of the problem through model architecture. Launching a new user experience to replace an established product involves sparse behavioral signals, low-engagement cohorts, and unstable model performance. We argue that effective recommendations require the synergistic integration of feature engineering, model architecture, and stable model updating. We propose Trinity, a framework embodying this principle. Trinity extracts valuable information from existing scenarios while ensuring predictive effectiveness and accuracy in the new scenario. In this paper, we showcase Trinity applied to a billion-user Microsoft product transition. Both offline and online experiments demonstrate that our framework achieves substantial improvements in addressing the combined challenge of new users in new scenarios.
Paper Structure (16 sections, 2 figures, 2 tables, 1 algorithm)

This paper contains 16 sections, 2 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Main Framework
  3. Feature Engineering
  4. Model Architecture
  5. Scenario Knowledge Extractor
  6. User Profile Adapter
  7. Stability-Aware Model Updating
  8. Experiment
  9. Experimental Settings
  10. Datasets
  11. Baselines and variants
  12. Hyper-Parameter
  13. Evaluation Metrics
  14. Performance Evaluation
  15. Online A/B Testing
  16. ...and 1 more sections

Figures (2)

  • Figure 1: Microsoft MSN product migration from classic style (left) to copilot style (right).
  • Figure 2: The model structure. This model is a multi-task model, where $t_1, t_2, \dots, t_n$ represent multi-classified durations. Since duration is similar to clicks and due to space limitations, duration-related content will not be further elaborated.