BroadGen: A Framework for Generating Effective and Efficient Advertiser Broad Match Keyphrase Recommendations

TL;DR

This paper tackles the challenge of generating effective and efficient broad-match keyphrases for sponsored search by introducing BroadGen, a three-core, token-correspondence framework that leverages historical query data to produce broad-match keyphrases. It formalizes the problem as string clustering over buyer queries, using anchor and query similarity matrices to form clusters, and derives representative keyphrases through common-token analysis and topological token ordering. The authors provide a comprehensive evaluation framework (RRE and PTR) and demonstrate that BroadGen, especially with data augmentation, achieves higher recall and balanced precision offline, while delivering strong online performance with significant gains in impressions and clicks in production on eBay. The work advances practical non-exact-match recommendations with explainability and scalability, offering a viable path toward integrating neural enhancements and embedding-based inputs in future iterations.

Abstract

In the domain of sponsored search advertising, the focus of Keyphrase recommendation has largely been on exact match types, which pose issues such as high management expenses, limited targeting scope, and evolving search query patterns. Alternatives like Broad match types can alleviate certain drawbacks of exact matches but present challenges like poor targeting accuracy and minimal supervisory signals owing to limited advertiser usage. This research defines the criteria for an ideal broad match, emphasizing on both efficiency and effectiveness, ensuring that a significant portion of matched queries are relevant. We propose BroadGen, an innovative framework that recommends efficient and effective broad match keyphrases by utilizing historical search query data. Additionally, we demonstrate that BroadGen, through token correspondence modeling, maintains better query stability over time. BroadGen's capabilities allow it to serve daily, millions of sellers at eBay with over 2.5 billion items.

Figures (6)

• Figure 1: A view of buyer (a) and seller side (b) of the keyphrase recommendation system for eBay advertising (manual targeting).
• Figure 2: Architecture for matching seller's keyphrases with different match types to buyer queries.
• Figure 3: Example recommendation of keyphrases by the Transfusion using NER tags.
• Figure 4: Example construction of keyphrases by BroadGen from clusters of queries. Illustration shows one cluster per item.
• Figure 5: Comparing Precision, Recall and F1 for BroadGen(Aug) with increasing count of output keywords across the categories.