Contextual Advertising by Combining Relevance with Click Feedback
Contextual advertising supports much of the Webs ecosystem today. User experience and revenue (shared by the publisher ad the ad network) depend on the relevance of displayed ads to the page content. As with other document retrieval systems, relevance is provided by scoring the match between individual ads (documents) and the content of page where the ads are shown (query). In this paper show how this match can be improved signi.cantly by augmenting the ad-page scoring function with extra parameters from a logistic regression model on the words in the pages and ads. A key property of the proposed model is that can be mapped to standard cosine similarity matching is suitable for e.cient and scalable implementation over inverted indexes. The model parameter values are learnt from logs containing ad impressions and clicks, with shrinkage estimators being used to combat sparsity. To scale our computations to train on an extremely large training corpus consisting of several gigabytes of data, we parallelize our.tting algorithm in a Hadoop 10 framework. Experimental evaluation is provided showing improved click prediction over holdout set of impression and click events from a large scale real-world ad placement engine. Our best model achieves a 25％ lift in precision relative to a traditional information retrieval model which is based on cosine similarity, for recalling 10％ of the clicks in our test data.
Clickthrough rate Modeling Interaction Effects
Deepayan Chakrabarti Deepak Agarwal Vanja Josifovski
Yahoo! Research 701 First Ave Sunnyvale, CA 94089.
国际会议
第十七届国际万维网大会(the 17th International World Wide Web Conference)(WWW08)
北京
英文
2008-04-21（万方平台首次上网日期，不代表论文的发表时间）