arXiv 2511.00694
Taxonomy-based Negative Sampling In Personalized Semantic Search for E-commerce
By Uthman Jinadu, Siawpeng Er, et al.
Published 2025-11-01
Discussion
Read the public discussion and references gathered around this paper.
Large retail outlets offer products that may be domain-specific, and this requires having a model that can understand subtle differences in similar items. Sampling techniques used to train these models are most of the time, computationally expensive or logistically challenging. These models also do not factor in users' previous purchase patterns or behavior, thereby retrieving irrelevant items for them. We present a…