Personalized Interest Graphs for Theme-Driven User Behavior

Oded Zinman; Nazmul Chowdhury; Leandro Fiaschetti; Yuri M. Brovman; Guy Feigenblat; Yotam Eshel

doi:10.1145/3705328.3748133

Personalized Interest Graphs for Theme-Driven User Behavior

Oded Zinman, Nazmul Chowdhury, Leandro Fiaschetti, Yuri M. Brovman, Guy Feigenblat, Yotam Eshel

eBay Inc.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Many eBay users turn to our platform to pursue theme-centric interests that span diverse product categories - for example, a Star Wars fan might search for related video games, toys, memorabilia, and artwork. Existing recommendation systems, typically optimized for short-term engagement, often fail to surface cross-category items aligned with these deeper interests. We present an end-to-end recommendation framework built around a user-interest graph generated by LLM chain. The graph captures user preferences at multiple levels of granularity, enabling a balance between relevance-driven and serendipity-driven recommendations. The system has been deployed at scale, serving millions of users across billions of items. An online A/B test on the eBay homepage showed a significant improvement in engagement with previously unseen categories, alongside gains in purchases and buyer count.

Original language	English
Title of host publication	RecSys2025 - Proceedings of the 19th ACM Conference on Recommender Systems
Publisher	Association for Computing Machinery, Inc
Pages	1038-1041
Number of pages	4
ISBN (Electronic)	9798400713644
DOIs	https://doi.org/10.1145/3705328.3748133
State	Published - 7 Aug 2025
Externally published	Yes
Event	19th ACM Conference on Recommender Systems, RecSys 2025 - Prague, Czech Republic Duration: 22 Sep 2025 → 26 Sep 2025

Publication series

Name	RecSys2025 - Proceedings of the 19th ACM Conference on Recommender Systems

Conference

Conference	19th ACM Conference on Recommender Systems, RecSys 2025
Country/Territory	Czech Republic
City	Prague
Period	22/09/25 → 26/09/25

Bibliographical note

Publisher Copyright:
© 2025 Copyright held by the owner/author(s).

Keywords

Knowledge Graph
Large Language Models
Recommendation

Access to Document

10.1145/3705328.3748133

Cite this

Zinman, O., Chowdhury, N., Fiaschetti, L., Brovman, Y. M., Feigenblat, G., & Eshel, Y. (2025). Personalized Interest Graphs for Theme-Driven User Behavior. In RecSys2025 - Proceedings of the 19th ACM Conference on Recommender Systems (pp. 1038-1041). (RecSys2025 - Proceedings of the 19th ACM Conference on Recommender Systems). Association for Computing Machinery, Inc. https://doi.org/10.1145/3705328.3748133

@inproceedings{6bd163b6e1584a95ac3be344eecbf6c3,

title = "Personalized Interest Graphs for Theme-Driven User Behavior",

abstract = "Many eBay users turn to our platform to pursue theme-centric interests that span diverse product categories - for example, a Star Wars fan might search for related video games, toys, memorabilia, and artwork. Existing recommendation systems, typically optimized for short-term engagement, often fail to surface cross-category items aligned with these deeper interests. We present an end-to-end recommendation framework built around a user-interest graph generated by LLM chain. The graph captures user preferences at multiple levels of granularity, enabling a balance between relevance-driven and serendipity-driven recommendations. The system has been deployed at scale, serving millions of users across billions of items. An online A/B test on the eBay homepage showed a significant improvement in engagement with previously unseen categories, alongside gains in purchases and buyer count.",

keywords = "Knowledge Graph, Large Language Models, Recommendation",

author = "Oded Zinman and Nazmul Chowdhury and Leandro Fiaschetti and Brovman, \{Yuri M.\} and Guy Feigenblat and Yotam Eshel",

note = "Publisher Copyright: {\textcopyright} 2025 Copyright held by the owner/author(s).; 19th ACM Conference on Recommender Systems, RecSys 2025 ; Conference date: 22-09-2025 Through 26-09-2025",

year = "2025",

month = aug,

day = "7",

doi = "10.1145/3705328.3748133",

language = "אנגלית",

series = "RecSys2025 - Proceedings of the 19th ACM Conference on Recommender Systems",

publisher = "Association for Computing Machinery, Inc",

pages = "1038--1041",

booktitle = "RecSys2025 - Proceedings of the 19th ACM Conference on Recommender Systems",

}

Zinman, O, Chowdhury, N, Fiaschetti, L, Brovman, YM, Feigenblat, G & Eshel, Y 2025, Personalized Interest Graphs for Theme-Driven User Behavior. in RecSys2025 - Proceedings of the 19th ACM Conference on Recommender Systems. RecSys2025 - Proceedings of the 19th ACM Conference on Recommender Systems, Association for Computing Machinery, Inc, pp. 1038-1041, 19th ACM Conference on Recommender Systems, RecSys 2025, Prague, Czech Republic, 22/09/25. https://doi.org/10.1145/3705328.3748133

Personalized Interest Graphs for Theme-Driven User Behavior. / Zinman, Oded; Chowdhury, Nazmul; Fiaschetti, Leandro et al.
RecSys2025 - Proceedings of the 19th ACM Conference on Recommender Systems. Association for Computing Machinery, Inc, 2025. p. 1038-1041 (RecSys2025 - Proceedings of the 19th ACM Conference on Recommender Systems).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Personalized Interest Graphs for Theme-Driven User Behavior

AU - Zinman, Oded

AU - Chowdhury, Nazmul

AU - Fiaschetti, Leandro

AU - Brovman, Yuri M.

AU - Feigenblat, Guy

AU - Eshel, Yotam

PY - 2025/8/7

Y1 - 2025/8/7

N2 - Many eBay users turn to our platform to pursue theme-centric interests that span diverse product categories - for example, a Star Wars fan might search for related video games, toys, memorabilia, and artwork. Existing recommendation systems, typically optimized for short-term engagement, often fail to surface cross-category items aligned with these deeper interests. We present an end-to-end recommendation framework built around a user-interest graph generated by LLM chain. The graph captures user preferences at multiple levels of granularity, enabling a balance between relevance-driven and serendipity-driven recommendations. The system has been deployed at scale, serving millions of users across billions of items. An online A/B test on the eBay homepage showed a significant improvement in engagement with previously unseen categories, alongside gains in purchases and buyer count.

AB - Many eBay users turn to our platform to pursue theme-centric interests that span diverse product categories - for example, a Star Wars fan might search for related video games, toys, memorabilia, and artwork. Existing recommendation systems, typically optimized for short-term engagement, often fail to surface cross-category items aligned with these deeper interests. We present an end-to-end recommendation framework built around a user-interest graph generated by LLM chain. The graph captures user preferences at multiple levels of granularity, enabling a balance between relevance-driven and serendipity-driven recommendations. The system has been deployed at scale, serving millions of users across billions of items. An online A/B test on the eBay homepage showed a significant improvement in engagement with previously unseen categories, alongside gains in purchases and buyer count.

KW - Knowledge Graph

KW - Large Language Models

KW - Recommendation

UR - https://www.scopus.com/pages/publications/105019645323

U2 - 10.1145/3705328.3748133

DO - 10.1145/3705328.3748133

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:105019645323

T3 - RecSys2025 - Proceedings of the 19th ACM Conference on Recommender Systems

SP - 1038

EP - 1041

BT - RecSys2025 - Proceedings of the 19th ACM Conference on Recommender Systems

PB - Association for Computing Machinery, Inc

T2 - 19th ACM Conference on Recommender Systems, RecSys 2025

Y2 - 22 September 2025 through 26 September 2025

ER -

Zinman O, Chowdhury N, Fiaschetti L, Brovman YM, Feigenblat G, Eshel Y. Personalized Interest Graphs for Theme-Driven User Behavior. In RecSys2025 - Proceedings of the 19th ACM Conference on Recommender Systems. Association for Computing Machinery, Inc. 2025. p. 1038-1041. (RecSys2025 - Proceedings of the 19th ACM Conference on Recommender Systems). doi: 10.1145/3705328.3748133

Personalized Interest Graphs for Theme-Driven User Behavior

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this