Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Survey for Landing Generative AI in Social and E-commerce Recsys -- the Industry Perspectives

Da Xu, Danqing Zhang, Guangyu Yang, Bo Yang, Shuyuan Xu, Lingling Zheng, Cindy Liang

TL;DR

Industrial Recommender Systems face substantive challenges in deploying Generative AI due to complex production infrastructures and operation requirements. The paper presents an application-centric survey of system foundations, LLMOps, RAG-based curation, AI agents for interactive recommendations, and responsible AI practices, linking practical engineering with current research advances. It offers a taxonomy, production-ready workflows, and open problems, aiming to accelerate real-world adoption while maintaining safety and alignment. By bridging industry experiences with research directions, the work provides concrete roadmaps for integrating GAI into social and e-commerce Recsys with tangible impact on user satisfaction and business goals.

Abstract

Recently, generative AI (GAI), with their emerging capabilities, have presented unique opportunities for augmenting and revolutionizing industrial recommender systems (Recsys). Despite growing research efforts at the intersection of these fields, the integration of GAI into industrial Recsys remains in its infancy, largely due to the intricate nature of modern industrial Recsys infrastructure, operations, and product sophistication. Drawing upon our experiences in successfully integrating GAI into several major social and e-commerce platforms, this survey aims to comprehensively examine the underlying system and AI foundations, solution frameworks, connections to key research advancements, as well as summarize the practical insights and challenges encountered in the endeavor to integrate GAI into industrial Recsys. As pioneering work in this domain, we hope outline the representative developments of relevant fields, shed lights on practical GAI adoptions in the industry, and motivate future research.

Survey for Landing Generative AI in Social and E-commerce Recsys -- the Industry Perspectives

TL;DR

Industrial Recommender Systems face substantive challenges in deploying Generative AI due to complex production infrastructures and operation requirements. The paper presents an application-centric survey of system foundations, LLMOps, RAG-based curation, AI agents for interactive recommendations, and responsible AI practices, linking practical engineering with current research advances. It offers a taxonomy, production-ready workflows, and open problems, aiming to accelerate real-world adoption while maintaining safety and alignment. By bridging industry experiences with research directions, the work provides concrete roadmaps for integrating GAI into social and e-commerce Recsys with tangible impact on user satisfaction and business goals.

Abstract

Recently, generative AI (GAI), with their emerging capabilities, have presented unique opportunities for augmenting and revolutionizing industrial recommender systems (Recsys). Despite growing research efforts at the intersection of these fields, the integration of GAI into industrial Recsys remains in its infancy, largely due to the intricate nature of modern industrial Recsys infrastructure, operations, and product sophistication. Drawing upon our experiences in successfully integrating GAI into several major social and e-commerce platforms, this survey aims to comprehensively examine the underlying system and AI foundations, solution frameworks, connections to key research advancements, as well as summarize the practical insights and challenges encountered in the endeavor to integrate GAI into industrial Recsys. As pioneering work in this domain, we hope outline the representative developments of relevant fields, shed lights on practical GAI adoptions in the industry, and motivate future research.
Paper Structure (31 sections, 15 figures)

This paper contains 31 sections, 15 figures.

Table of Contents

  1. Introduction
  2. Outline
  3. Key Contributions and Distinguishing from Related Works
  4. Background and Fundamentals
  5. Industrial Recsys in a Nutshell
  6. Major Deficiencies of Existing Systems
  7. Building GAI Foundation and LLMOps
  8. GAI ecosystem.
  9. LLMOps.
  10. Classification Framework of GAI in Recsys
  11. GAI-enhanced data, feature, and modeling for Recsys
  12. Research Progress
  13. Solution Framework and Practical Considerations
  14. Augmenting the Curation Capability of Recsys
  15. Research Progress
  16. ...and 16 more sections

Figures (15)

  • Figure 1: Structure of the survey.
  • Figure 2: Overview of the key components employed by modern industrial social and e-commerce Recsys. The modeling stack typically consists of data, representation, modeling, objective, and product layers. The infra stack often comprises diverse serving, streaming, backend serveries orchestrated using MLOps techniques.
  • Figure 3: Illustration of the GAI foundations that will be the focus areas of our tutorial.
  • Figure 4: Overview of DevOps, MLOps, and LLMOps. The key LLMOps terms will be clarified in the rest of this survey.
  • Figure 5: Real-world product taxonomy of social/e-commerce Recsys and they best fit into the major solution categories of GAI in Recsys solutions.
  • ...and 10 more figures