One step ad retrieval improves real time search performance
One-Step Retrieval Framework for Real-Time Sponsored Search Ads Using Hierarchical Text Representations
Information Retrieval
Summary
Online ads are usually picked for searches using complicated multi-step systems that can miss good matches. The authors found these old ways also struggle to handle new ads well and can be slow. They created a new method called ANGLE that uses a large AI language model to understand ads and search queries in a unified way, scoring ads for relevance and value all at once. This approach showed better results in live tests with more sales and improved ad matching accuracy in offline tests.
What this means in practice
- •For online advertising teams: Improve how search ads are selected and ranked in real time by using a unified AI model that better understands ad text and intent.$Commercial implications: Enables development of more effective real-time ad selection and ranking products that increase sales and customer engagement.
- •For search engine engineers: Update retrieval architecture to replace multi-stage cascades with single-step unified models for faster and more accurate ad retrieval.
Authors
Tongtong Liu, Renyu Zhang, Jiayu Ding, Hongchao Guo, Xintao Yang, He Wei, Zhaoyu Li, Haiyang Wu
Abstract
Traditional retrieval systems typically use multi-stage cascading architectures (MCA), where each module is optimized independently, leading to inconsistent objectives and the premature elimination of high-potential candidates. Recent LLM-based generation methods offer end-to-end solutions but use discrete semantic identifiers (SIDs) to retrieve ads, which are not learned by the base LLM and require memorization of numerous SID-to-ad mappings during SFT, suffering from limited generalization to unseen ads, high maintenance and update costs. The one-to-one mapping between SIDs and advertisements leads to inefficient decoding. Moreover, these methods rely on a small reward model (e.g. pctr) for relevance and ranking, limiting the LLM's ability to fully assess ads' commercial value. To address these challenges, we propose A uNified Generation-discriminative-ranking reaL-time rEtrieval (ANGLE) framework. ANGLE uses LLM-generated hierarchical textual representations, which consist of commercial intent that provide high-level overviews and ad abstract that deliver fine-grained details. Additionally, ANGLE integrates retrieval, relevance, and ranking directly within a single LLM, enabling precise and efficient ranking of ads by leveraging the full capabilities of the LLM. We applied ANGLE to the real-world search scenarios, achieving a 1.81% increase in consumption and a 2.16% increase in gross merchandise volume (GMV). We also conducted offline evaluations of ANGLE and seven baselines, with ANGLE outperforming all across key metrics such as HR and ACR.