Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Robust Interaction-Based Relevance Modeling for Online e-Commerce Search

Ben Chen, Huangyu Dai, Xiang Ma, Wen Jiang, Wei Ning

TL;DR

This work tackles semantic relevance calculation for large-scale online e-commerce search by proposing a robust interaction-based model (ei-SRC). It introduces three innovations—dynamic-length representations to reduce inference cost, domain-specific vocabulary expansion with professional-terms recognition, and a contrastive adversarial training (CAT) framework to enhance robustness and matching accuracy. Offline evaluations show ei-SRC outperforms strong baselines and benefits from ablation-enabled cost reductions, while online A/B tests on Alibaba.com demonstrate meaningful gains in CTR, CVR, and PAY, culminating in deployment across full daily traffic. The results suggest that a lean, interaction-based architecture can achieve superior relevance under strict latency constraints and deliver tangible business value at scale.

Abstract

Semantic relevance calculation is crucial for e-commerce search engines, as it ensures that the items selected closely align with customer intent. Inadequate attention to this aspect can detrimentally affect user experience and engagement. Traditional text-matching techniques are prevalent but often fail to capture the nuances of search intent accurately, so neural networks now have become a preferred solution to processing such complex text matching. Existing methods predominantly employ representation-based architectures, which strike a balance between high traffic capacity and low latency. However, they exhibit significant shortcomings in generalization and robustness when compared to interaction-based architectures. In this work, we introduce a robust interaction-based modeling paradigm to address these shortcomings. It encompasses 1) a dynamic length representation scheme for expedited inference, 2) a professional terms recognition method to identify subjects and core attributes from complex sentence structures, and 3) a contrastive adversarial training protocol to bolster the model's robustness and matching capabilities. Extensive offline evaluations demonstrate the superior robustness and effectiveness of our approach, and online A/B testing confirms its ability to improve relevance in the same exposure position, resulting in more clicks and conversions. To the best of our knowledge, this method is the first interaction-based approach for large e-commerce search relevance calculation. Notably, we have deployed it for the entire search traffic on alibaba.com, the largest B2B e-commerce platform in the world.

Robust Interaction-Based Relevance Modeling for Online e-Commerce Search

TL;DR

This work tackles semantic relevance calculation for large-scale online e-commerce search by proposing a robust interaction-based model (ei-SRC). It introduces three innovations—dynamic-length representations to reduce inference cost, domain-specific vocabulary expansion with professional-terms recognition, and a contrastive adversarial training (CAT) framework to enhance robustness and matching accuracy. Offline evaluations show ei-SRC outperforms strong baselines and benefits from ablation-enabled cost reductions, while online A/B tests on Alibaba.com demonstrate meaningful gains in CTR, CVR, and PAY, culminating in deployment across full daily traffic. The results suggest that a lean, interaction-based architecture can achieve superior relevance under strict latency constraints and deliver tangible business value at scale.

Abstract

Semantic relevance calculation is crucial for e-commerce search engines, as it ensures that the items selected closely align with customer intent. Inadequate attention to this aspect can detrimentally affect user experience and engagement. Traditional text-matching techniques are prevalent but often fail to capture the nuances of search intent accurately, so neural networks now have become a preferred solution to processing such complex text matching. Existing methods predominantly employ representation-based architectures, which strike a balance between high traffic capacity and low latency. However, they exhibit significant shortcomings in generalization and robustness when compared to interaction-based architectures. In this work, we introduce a robust interaction-based modeling paradigm to address these shortcomings. It encompasses 1) a dynamic length representation scheme for expedited inference, 2) a professional terms recognition method to identify subjects and core attributes from complex sentence structures, and 3) a contrastive adversarial training protocol to bolster the model's robustness and matching capabilities. Extensive offline evaluations demonstrate the superior robustness and effectiveness of our approach, and online A/B testing confirms its ability to improve relevance in the same exposure position, resulting in more clicks and conversions. To the best of our knowledge, this method is the first interaction-based approach for large e-commerce search relevance calculation. Notably, we have deployed it for the entire search traffic on alibaba.com, the largest B2B e-commerce platform in the world.
Paper Structure (10 sections, 8 equations, 2 figures, 7 tables)

This paper contains 10 sections, 8 equations, 2 figures, 7 tables.

Table of Contents

  1. Introduction
  2. related works
  3. methodology
  4. Dynamic-length Representation Scheme
  5. Professional Terms Recognition Strategy
  6. Contrastive Adversarial Training Mechanism
  7. experiments
  8. Offline evaluation.
  9. Online Experiments.
  10. Conclusion

Figures (2)

  • Figure 1: The overview of the ei-SRC, the proposed e-commerce interaction-based semantic relevance calculation method. For brevity, the dynamic-length representation scheme is not illustrated, which will be displayed separately.
  • Figure 2: The overview of dynamic-length representation scheme, where each square is a token. The number shown on it simulates the sum of various types of embeddings.