Personalized Re-ranking for Recommendation with Mask Pretraining

Data Science and Engineering

Research Paper

Zichen Xu¹  Peng Han²  Silin Zhou²  Jie Yu²  Shuo Shang²  Lisi Chen² 

[1]Nanchang University, Nanchang, China

[2]University of Electronic Science and Technology of China, Chengdu, China

关键词: Re-ranking;    Recommendation;    Mask pretraining;    Scene;   

DOI  :  10.1007/s41019-023-00219-6

 received in 2023-06-11, accepted in 2023-07-24,  发布年份 2023

来源: Springer

【 摘 要 】

Re-ranking is to refine the candidate ranking list of recommended items, such that the re-ranked list attracts users to purchase or click more items than the candidate one without re-ranking. Items in the candidate list are often ranked by their relevance to users’ interests. It is thus important to exploit the mutual influence between items in the re-ranking process. Existing re-ranking models focus on only the pairwise influence between two items, and have limited capability to exploit the local mutual influence in a group of items. Users often show successive interests on a group of relevant items, e.g., mobile phone, phone covers, wireless headset, namely scene. We propose a novel re-ranking model that jointly exploits the local mutual influence in scenes and the global mutual influence between different scenes. Scene representations are learned by GNN and multi-head attention, where GNN aims to learn local mutual influence while multi-head attention is to learn global mutual influence. To study the interaction between users and scenes, matrix factorization on users is utilized to obtain the user preference, which can be further applied to scenes to compute the scene scores. The final re-ranking list is generated by sorting the predicted scores of all scenes. To further mine user history information and item related user information, we also develop the extension pretraining module which relies on mask mechanism to support users and items high-quality embedding generation. We conduct a comprehensive evaluation on several real-world datasets. The experimental results demonstrate that our model substantially outperforms existing approaches.

【 授权许可】

CC BY   
© The Author(s) 2023

【 图 表 】

12951_2015_155_Article_IEq32.gif

Fig. 1

12864_2016_2889_Article_IEq3.gif

Fig. 1

Fig. 2

Fig. 2

Fig. 2

Fig. 1

12951_2017_270_Article_IEq10.gif

Fig. 8

Fig. 3

Fig. 3

Fig. 7

Fig. 1

12951_2015_155_Article_IEq30.gif

Fig. 5

Fig. 6

12951_2017_270_Article_IEq9.gif

Fig. 2

12951_2017_270_Article_IEq8.gif

Fig. 1

12951_2015_111_Article_IEq4.gif

Fig. 1

13100_2023_302_Article_IEq7.gif

Fig. 2

Fig. 4

Fig. 5

12951_2015_155_Article_IEq28.gif

42004_2023_1020_Article_IEq18.gif

Fig. 3

Fig. 7

13100_2023_302_Article_IEq2.gif

13100_2023_302_Article_IEq1.gif

12951_2017_270_Article_IEq7.gif

42004_2023_1020_Article_IEq12.gif

Fig. 2

Fig. 7

Fig. 1

Fig. 3

12951_2015_111_Article_IEq2.gif

Fig. 1

Fig. 5

Fig. 6

12951_2015_155_Article_IEq27.gif

12951_2017_315_Article_IEq2.gif

12951_2015_155_Article_IEq26.gif

12951_2015_155_Article_IEq25.gif

12951_2015_155_Article_IEq24.gif

Fig. 2

12951_2017_270_Article_IEq5.gif

Fig. 3

Fig. 1

12951_2015_111_Article_IEq1.gif

Fig. 6

Fig. 4

Fig. 1

12951_2017_315_Article_IEq1.gif

Fig. 1

12951_2015_155_Article_IEq23.gif

12951_2015_155_Article_IEq22.gif

12951_2015_155_Article_IEq21.gif

12951_2015_155_Article_IEq20.gif

12951_2015_155_Article_IEq19.gif

12951_2015_155_Article_IEq16.gif

Fig. 3

12951_2015_155_Article_IEq14.gif

12951_2015_155_Article_IEq13.gif

12951_2015_155_Article_IEq12.gif

Fig. 2

12951_2015_155_Article_IEq9.gif

Fig. 2

12864_2017_3527_Article_IEq9.gif

Fig. 7

Fig. 2

Fig. 3

Fig. 5

Fig. 5

Fig. 10

Fig. 1

Fig. 9

12951_2015_155_Article_IEq7.gif

Fig. 8

Fig. 4

Fig. 3

Fig. 2

Fig. 4

Fig. 1

Fig. 3

Fig. 1

40517_2023_273_Article_IEq9.gif

Fig. 2

Fig. 2

Fig. 4