Predicting ads' quality
Internet advertising is the main source of income for search engines today. As the number of Internet users increases, the Internet advertising becomes increasingly popular among people who want to advertise a service or a product. Google reported $6,475 million revenue from advertisement in 2009 which is 8% more than the previous year. This, emphasizes the fact that internet advertising is a widely attractive and growing market for advertisers and search engines.
When a user enters a query in a search engine, there are often some sponsored links or ads presented alongside with the search results. These ads are chosen by an auction between all candidate ads which have keywords similar to the user entered query. In this auction, winners will be chosen based on two factors: offered bid and quality. In this article, quality means the ability to attract more users' clicks. Advertisers usually want to place their ads in the best spot in the page without paying more money, so they try to increase the quality of ads by choosing good terms for title and descriptions. On the other side, Search engines use the Price Per Click (PPC) model for Internet advertising. In this model search engines can earn money just if somebody clicks on the displayed ads and as a result, there is no cost for the advertisers merely because of ad appearance. So for earning maximum revenue, search engines also try to select ads with better quality to attract more clicks. Roughly speaking, ads with high quality are important for both advertiser and search engine.
For the ads which have been in the system for longer periods of the time, we can find their quality just by looking at their click through rate (CTR). If an ad had higher amount of CTR, it is more attractive to users and has better quality. But for new ads or for those ones without enough historical data, we should find another way to estimate their quality.
People
On-going Research Problems
Estimating the click-through rate for new ads with semantic and feature based similarity algorithms
References and Links
some good references are available here
