Full Paper (10 pages)
Official ACM published version: http://dx.doi.org/10.1145/2556195.2556257
Author's version: PDF (291KB)
In modern commercial search engines, the pay-per-click (PPC) advertising model is widely used in sponsored search. The search engines try to deliver ads which can produce greater click yields (the total number of clicks for the list of ads per impression). Therefore, predicting user clicks plays a critical role in sponsored search. The current ad-delivery strategy is a two-step approach which first predicts individual ad CTR for the given query and then selects the ads with higher predicted CTR. However, this strategy is naturally suboptimal and correlation between ads is often ignored under this strategy. The learning problem is focused on predicting individual performance rather than group performance which is the more important measurement.
In this paper, we study click yield measurement in sponsored search and focus on the problem-predicting group performance (click yields) in sponsored search. To tackle all challenges in this problem--depth effects, interactive influence, cold start and sparseness of ad textual information--we first investigate several effects and propose a novel framework that could directly predict group performance for lists of ads. Our extensive experiments on a large-scale real-world dataset from a commercial search engine show that we achieve significant improvement by solving the sponsored search problem from the new perspective. Our methods noticeably outperform existing state-of-the-art approaches.
In Proceedings of the 7th Annual ACM International Conference on Web Search and Data Mining (WSDM), pages 143-152, New York City, NY, February 2014.
Back to Brian Davison's publications