Learning to Display in Sponsored Search
In sponsored search, it is necessary for the search engine, to decide the right number of advertisements (ads) to display for each query, in the constraint of a limited commercial load. Because over displaying ads will lead to the commercial overload problem, driving some of the users away in the long run. Despite the importance of the issue, very few literatures have discussed about how to measure the commercial load in sponsored search. Thus it is difficult for the search engine to make decisions quantitatively in practice. As a primary study, we propose to quantify the commercial load by the average displayed ad number per query, and then we investigate the displaying strategy to optimize the total revenue, in the constraint of a limited commercial load. We formalize this task under the framework of the secretary problem. A novel dynamic algorithm is proposed, which is extended from the state-of-the-art multiple-choice secretary algorithm. Through theoretical analysis, we proof that our algorithm is approaching the optimal value; and through empirical analysis, we demonstrate that our algorithm outperforms the fundamental static algorithm significantly. The algorithm can scale up with respect to very large datasets.
The work described in this paper was fully supported by National Natural Science Foundation of China (No. 61300076) and Ph.D. Programs Foundation of Ministry of Education of China (No. 20131101120035).
- 1.Broder, A., Ciaramita, M., Fontoura, M., Gabrilovich, E., Josifovski, V., Metzler, D., Murdock, V., Plachouras, V.: To swing or not to swing: learning when (not) to advertise. In: Proceedings of CIKM’08, pp. 1003–1012. ACM (2008)Google Scholar
- 2.Buscher, G., Dumais, S.T., Cutrell, E.: The good, the bad, and the random: an eye-tracking study of ad quality in web search. In: Proceedings of SIGIR’10, pp. 42–49. ACM (2010)Google Scholar
- 3.Graepel, T., Candela, J.Q., Borchert, T., Herbrich, R.: Web-scale bayesian click-through rate prediction for sponsored search advertising in microsofts bing search engine. In: Proceedings of ICML’10 (2010)Google Scholar
- 4.Guo, F., Liu, C., Kannan, A., Minka, T., Taylor, M., Wang, Y.M., Faloutsos, C.: Click chain model in web search. In: Proceedings of WWW’09, pp. 11–20. ACM (2009)Google Scholar
- 6.Kleinberg, R.: A multiple-choice secretary algorithm with applications to online auctions. In: Proceedings of ACM-SIAM Symposium on Discrete algorithms, pp. 630–631 (2005)Google Scholar
- 8.Marable, L.: False oracles: consumer reaction to learning the truth about how search engines work, results of an ethnographic study (2003)Google Scholar
- 9.Nath, A., Mukherjee, S., Jain, P., Goyal, N., Laxman, S.: Ad impression forecasting for sponsored search. In: Proceedings of WWW’13, pp. 943–952. ACM (2013)Google Scholar
- 10.Pandey, S., Punera, K., Fontoura, M., Josifovski, V.: Estimating advertisability of tail queries for sponsored search. In: Proceedings of SIGIR’10, pp. 563–570 (2010)Google Scholar
- 11.Radovanovic, A., Heavlin, W.D.: Risk-aware revenue maximization in display advertising. In: Proceedings of WWW’12, pp. 91–100. ACM (2012)Google Scholar
- 12.Zhu, Y., Wang, G., Yang, J., Wang, D., Yan, J., Hu, J., Chen, Z.: Optimizing search engine revenue in sponsored search. In Proceedings of SIGIR’09, pp. 588–595. ACM (2009)Google Scholar