Yahoo! Research Talk
Hema Raghavan (UMass Amherst CS Ph.D. '07)
Research Scientist, Yahoo! Applied Research
Friday, October 10, 2008
9:15 a.m. in CS 151
"Predicting the Click-Through-Rate of Advertisements for Sponsored Search"
ABSTRACT: The primary source of revenue for major search engines is through advertising. In the popular auction model used by search engines, advertisers bid on queries or phrases that users may type in when looking for a product or information related to the advertisement's content. When a user types a query, advertisements that have bidded-phrases associated with the user's query are shown. The advertiser pays the search engine an amount determined by the second-price auction model if and only if the user clicks on the ad. This model allows advertisers to target ads to interested users only. Search engines popularly rank ads by a combination of bid and relevance in the hope of maximizing revenue. In this talk we will focus on the problem of determining relevance of an ad to a user query, and introduce several interesting areas of research that our group at Yahoo! is involved in. I will then describe in detail one aspect of our work where we build models to predict the click through rate of an ad. Our models allow for implicit query expansion via translation tables derived from multiple sources and allow us to directly estimate the probability that an advertisement is clicked on for a given query. We evaluate several variants of popular information retrieval models for the problem of Sponsored Search. Our models are estimated from click-through-data and are general enough that they can be applied to a wide variety of domains.
Refreshments provided prior to the talk at 9:00 a.m.