Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Beyond Positive History: Re-ranking with List-level Hybrid Feedback

Muyan Weng, Yunjia Xi, Weiwen Liu, Bo Chen, Jianghao Lin, Ruiming Tang, Weinan Zhang, Yong Yu

TL;DR

This work proposes Re-ranking with List-level Hybrid Feedback, a re-ranking system that captures user's preferences and behavior patterns with three modules: a Disentangled Interest Miner to disentangle the user's preferences into interests and disinterests, a Sequential Preference Mixer to learn users' entangled preferences considering the context of feedback, and a Comparison-aware Pattern Extractor to capture user's behavior patterns within each list.

Abstract

As the last stage of recommender systems, re-ranking generates a re-ordered list that aligns with the user's preference. However, previous works generally focus on item-level positive feedback as history (e.g., only clicked items) and ignore that users provide positive or negative feedback on items in the entire list. This list-level hybrid feedback can reveal users' holistic preferences and reflect users' comparison behavior patterns manifesting within a list. Such patterns could predict user behaviors on candidate lists, thus aiding better re-ranking. Despite appealing benefits, extracting and integrating preferences and behavior patterns from list-level hybrid feedback into re-ranking multiple items remains challenging. To this end, we propose Re-ranking with List-level Hybrid Feedback (dubbed RELIFE). It captures user's preferences and behavior patterns with three modules: a Disentangled Interest Miner to disentangle the user's preferences into interests and disinterests, a Sequential Preference Mixer to learn users' entangled preferences considering the context of feedback, and a Comparison-aware Pattern Extractor to capture user's behavior patterns within each list. Moreover, for better integration of patterns, contrastive learning is adopted to align the behavior patterns of candidate and historical lists. Extensive experiments show that RELIFE significantly outperforms SOTA re-ranking baselines.

Beyond Positive History: Re-ranking with List-level Hybrid Feedback

TL;DR

This work proposes Re-ranking with List-level Hybrid Feedback, a re-ranking system that captures user's preferences and behavior patterns with three modules: a Disentangled Interest Miner to disentangle the user's preferences into interests and disinterests, a Sequential Preference Mixer to learn users' entangled preferences considering the context of feedback, and a Comparison-aware Pattern Extractor to capture user's behavior patterns within each list.

Abstract

As the last stage of recommender systems, re-ranking generates a re-ordered list that aligns with the user's preference. However, previous works generally focus on item-level positive feedback as history (e.g., only clicked items) and ignore that users provide positive or negative feedback on items in the entire list. This list-level hybrid feedback can reveal users' holistic preferences and reflect users' comparison behavior patterns manifesting within a list. Such patterns could predict user behaviors on candidate lists, thus aiding better re-ranking. Despite appealing benefits, extracting and integrating preferences and behavior patterns from list-level hybrid feedback into re-ranking multiple items remains challenging. To this end, we propose Re-ranking with List-level Hybrid Feedback (dubbed RELIFE). It captures user's preferences and behavior patterns with three modules: a Disentangled Interest Miner to disentangle the user's preferences into interests and disinterests, a Sequential Preference Mixer to learn users' entangled preferences considering the context of feedback, and a Comparison-aware Pattern Extractor to capture user's behavior patterns within each list. Moreover, for better integration of patterns, contrastive learning is adopted to align the behavior patterns of candidate and historical lists. Extensive experiments show that RELIFE significantly outperforms SOTA re-ranking baselines.

Paper Structure

This paper contains 26 sections, 15 equations, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminaries
  4. Methodology
  5. Overview
  6. Embedding & Intra-Candidate Context (ICC)
  7. Disentangled Interest Miner (DIM)
  8. Sequential Preference Mixer (SPM)
  9. Comparison-aware Pattern Extractor (CPE)
  10. Prediction and Optimization
  11. Complexity Analysis
  12. Experiment
  13. Setup
  14. Datasets
  15. Baselines
  16. ...and 11 more sections

Figures (5)

  • Figure 1: Comparison between re-ranking with positive feedback (above) and list-level hybrid feedback (below).
  • Figure 2: The overall framework of RELIFE mainly consists of Intra-Candidate Context (ICC), Disentangled Interest Miner (DIM), Sequential Preference Mixer (SPM), and Comparison-aware Pattern Extractor (CPE).
  • Figure 3: The impact of the number of history lists.
  • Figure 4: The impact of the weight $\beta$ of the InfoNCE loss.
  • Figure 5: Similarity between positive and negative item embeddings from MIR and RELIFE. $\blacktriangle$ denotes positive candidate items, $\blacktriangle$ denotes positive historical items. $\bigstar$ denotes negative candidate items, $\bigstar$ denotes negative historical items.