Observations and research on the Internet online advertisement





吉田キャンパス 工学部10号館1階124号 第一講義室


Min Zhang 准教授


中国 清華大学准教授
(Associate Professor of Department of Computer Science and Technology at the Tsinghua University, China)


Dr. Min Zhang is an associate professor in the Department of Computer Science & Technology (DCST), Tsinghua University. She received her Bachelor and PhD degrees from DCST at Tsinghua University in 1999 and 2003, respectively. Dr. Zhang specializes in information retrieval, Web user behavior analysis and machine learning. She has published more than 50 papers on journals and conferences, such as JASIST, IR, SIGIR, WWW, CIKM, etc. She has participated in TREC (Text REtrieval Conference) as the team leader since 2002. In TREC benchmarks, her team has continuously achieved multiple top performances during the past 8 years. Dr. Zhang also teaches the courses of introduction for machine learning and advanced topics in information retrieval, and has achieved Best Teaching Activities Award for Young Faculties in Tsinghua University.


Observations and research on the Internet online advertisement


With the explosive growth of Web information and user activity, WWW becomes one of the most important media for advertisers. Derived from the two main users' information access ways, namely search and navigation, sponsored search and contextual advertising have been two most popular manners for the state-of-the-art Internet advertising. In this talk, I will discuss our observations on the users' interaction behavior with sponsored search, such as such as: Does advertising links in search result list result in bad user experience? How effective are these advertising links? How does user's click behavior change with different search queries and different ads positions?
The study was made on the large scale commercial search engine user log analysis. On contextual advertising, a novel contextual advertising mechanism based on ads and Web page classification techniques will be introduced. Experimental results on the practical commercial ads data prove the effectiveness of the proposed approach.