Event

Clickthrough Rate Prediction for Sponsored Search Advertising

Ozgur Cetin

Yahoo!

Tuesday, October 21, 2008
12:30

Sponsored search advertising where the ads are displayed in response the user search queries has become one of the biggest financial applications of the web. In the first part of this talk, I will give an overview of sponsored search advertising, and some of the machine learning, information retrieval, and optimization problems involved. One of the key problems in sponsored search advertising is clickthrough rate (CTR) prediction for the query-ad pairs. Accurate CTR prediction allows for displaying most relevant ads to queries, improving both user experience and search engine revenue and advertiser conversion. In the second part of the talk, I will talk about some of the recent feature extraction and statistical modeling research performed at Yahoo!, for accurate CTR prediction. For feature extraction, I will present methods to utilize information in user logs and advertiser texts. For statistical modeling, I will present a maximum entropy-based approach, and how one can do unsupervised adaptation to query clusters or groups of users in that framework.