Descripció del projecte
Online advertising is the main current revenue model on the Internet, hence making possible most of the freely available content and services [1]. Studies suggest that if online ads provide quality information about a product, a customer would rather prefer stop looking for other (and potentially better) options to buy and stick to the suggested ones [2]. Thus, advertisers are interested in serving the most beneficial advertisement (also known as ads) to the most relevant customer to have a high return-on-investment (ROI) [3], [4]. However, given the high complexity of this ecosystem, campaign optimization for higher conversions is not trivial and has become one of the major industry challenges. On the other side, publishers seek to earn the maximum revenue by selling their site space to advertisers, but at the same time they need to attract and retain high-quality traffic so that their sites can bring profits as well.
This ecosystem has an important and direct impact on web-users. Beside privacy issues [5], some ads might also lead to negative experiences such as failing to reach real content, confusion, tiredness, which can contribute to develop negative attitudes towards websites hosting ads [6]. Because of this, many users decide to install ad-blocking software to make their navigation more pleasant [7]. However, well designed ads can also impact positively on people by providing help exploring products that might turn out to be interesting or useful for them.
The purpose of this PhD thesis is to contribute to the improvement of the online advertising ecosystem by proposing methods that can benefit three of the main stakeholders involved in this ecosystem: publishers, advertisers and final users receiving the ads. The main contributions will be to develop tools for the detection of suspicious and misleading ad content using computer vision, improve the viewability of ads by learning user behavior from previous data and enhance the matchmaking between ads and users using machine learning techniques. All these projects can have a very beneficial impact on the advertising ecosystem and their actors. The research conducted for this PhD thesis will be performed within ExoClick’s ad-network company with real data of ads and traffic. ExoClick is ranked the 4th largest ad network in the world, providing a great testbed with a daily average of 6 billion impressions and more than 20 different ad types reaching most countries and users in the world.
1. Evans, D.S., 2009. The online advertising industry: Economics, evolution, and privacy. Journal of Economic Perspectives, 23(3), pp.37-60.
2. Ward, M.R. and Lee, M.J., 2000. Internet shopping, consumer search and product branding. Journal of product & brand management, 9(1), pp.6-20.
3. Geyik, S.C., Shen, J., Shariat, S., Dasdan, A. and Kolay, S., 2017. Towards Data Quality Assessment in Online Advertising. arXiv preprint arXiv:1711.11175.
4. Karmaker Santu, S.K., Li, L., Park, D.H., Chang, Y. and Zhai, C., 2017, April. Modeling the Influence of Popular Trending Events on User Search Behavior. In Proceedings of the 26th International Conference on World Wide Web Companion, pp. 535-544.
5. Abbadi, A. El. 2009. Fraud , Anonymization , and Privacy in the Internet : A Database Perspective, pp.15–16.
6. Brajnik, G. and Gabrielli, S., 2010. A review of online advertising effects on the user experience. International Journal of Human-Computer Interaction, 26(10), pp.971-997.
7. Experience, A., Proposed, P., Ads, B., & Ad, R. (n.d.). Determining a Better Ads Standard Based on User Experience Data, pp.1–46.
