Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multi-Scenario User Profile Construction via Recommendation Lists

Hui Zhang, Jiayu Liu

Abstract

Recommender systems (RS) play a core role in various domains, including business analytics, helping users and companies make appropriate decisions. To optimize service quality, related technologies focus on constructing user profiles by analyzing users' historical behavior information. This paper considers four analytical scenarios to evaluate user profiling capabilities under different information conditions. A generic user attribute analysis framework named RAPI is proposed, which infers users' personal characteristics by exploiting easily accessible recommendation lists. Specifically, a surrogate recommendation model is established to simulate the original model, leveraging content embedding from a pre-trained BERT model to obtain item embeddings. A sample augmentation module generates extended recommendation lists by considering similarity between model outputs and item embeddings. Finally, an adaptive weight classification model assigns dynamic weights to facilitate user characteristic inference. Experiments on four collections show that RAPI achieves inference accuracy of 0.764 and 0.6477, respectively.

Multi-Scenario User Profile Construction via Recommendation Lists

Abstract

Recommender systems (RS) play a core role in various domains, including business analytics, helping users and companies make appropriate decisions. To optimize service quality, related technologies focus on constructing user profiles by analyzing users' historical behavior information. This paper considers four analytical scenarios to evaluate user profiling capabilities under different information conditions. A generic user attribute analysis framework named RAPI is proposed, which infers users' personal characteristics by exploiting easily accessible recommendation lists. Specifically, a surrogate recommendation model is established to simulate the original model, leveraging content embedding from a pre-trained BERT model to obtain item embeddings. A sample augmentation module generates extended recommendation lists by considering similarity between model outputs and item embeddings. Finally, an adaptive weight classification model assigns dynamic weights to facilitate user characteristic inference. Experiments on four collections show that RAPI achieves inference accuracy of 0.764 and 0.6477, respectively.
Paper Structure (33 sections, 18 equations, 6 figures, 12 tables, 1 algorithm)

This paper contains 33 sections, 18 equations, 6 figures, 12 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Collaborative Filtering
  4. User Profile Analysis in Recommender Systems
  5. Preliminary
  6. Notation Definition
  7. Scenario Introduction
  8. Scenario 1
  9. Scenario 2
  10. Scenario 3
  11. Scenario 4
  12. Method
  13. Spurious Model Confirmation
  14. RLDA: Data Augmentation for Recommendation List
  15. Adaptive Weight Classification Model
  16. ...and 18 more sections

Figures (6)

  • Figure 1: The illustration of four scenarios
  • Figure 2: The overall framework of our proposed RAPI. The upper part of the figure illustrates the analyst's actions for characterizing user attributes in Scenario 1 and 2. The lower part of the figure illustrates the analyst's actions for characterizing user attributes in Scenario 3 and 4.
  • Figure 3: Statistics about sensitive attributes in records.
  • Figure 4: Recommendation performance of various candidate models.
  • Figure 5: Impact of the aggregation function on inference accuracy.
  • ...and 1 more figures