Semantic Ads Retrieval at Walmart eCommerce with Language Models Progressively Trained on Multiple Knowledge Domains

TL;DR

Sponsored search retrieval in Walmart's e-commerce context faces challenges of query–product misalignment, intent ambiguity, and sparse data. The authors present an end-to-end semantic retrieval pipeline built around a BERT-based model pretrained on Walmart product categories, a two-tower Siamese embedding, and a progressive fusion training regime with human-in-the-loop weighting and hard negative sampling. They demonstrate substantial offline gains (up to 16% NDCG) and online revenue improvements after production deployment in 2023, validating the approach's practicality at scale. This work offers a production-ready framework for semantic ads retrieval that blends domain-specific knowledge with neural retrieval to boost relevance and business metrics.

Abstract

Sponsored search in e-commerce poses several unique and complex challenges. These challenges stem from factors such as the asymmetric language structure between search queries and product names, the inherent ambiguity in user search intent, and the vast volume of sparse and imbalanced search corpus data. The role of the retrieval component within a sponsored search system is pivotal, serving as the initial step that directly affects the subsequent ranking and bidding systems. In this paper, we present an end-to-end solution tailored to optimize the ads retrieval system on Walmart.com. Our approach is to pretrain the BERT-like classification model with product category information, enhancing the model's understanding of Walmart product semantics. Second, we design a two-tower Siamese Network structure for embedding structures to augment training efficiency. Third, we introduce a Human-in-the-loop Progressive Fusion Training method to ensure robust model performance. Our results demonstrate the effectiveness of this pipeline. It enhances the search relevance metric by up to 16% compared to a baseline DSSM-based model. Moreover, our large-scale online A/B testing demonstrates that our approach surpasses the ad revenue of the existing production model.

Figures (6)

• Figure 1: Two-stage progressive training architecture of retrieval embedding model.
• Figure 2: Human-in-the-loop knowledge fusion sampling.
• Figure 3: Department classification label of exampled item
• Figure 4: Pipeline architecture of generating Ads embedding
• Figure 5: Retrieval service architecture based on Vespa search engine