Keywords:learning to rank, multimodal, deep learning
Online advertising platforms have a straight impact to be able to appeal various ad creatives to a significant amount of users on web pages and applications. They construct their own algorithms for ad optimization, so advertisers have tackled to improve ad performance. Creating a wide variety of ad creatives is the core task to get more feedback from users. However, we need to decide the priority as candidate ads because the platforms have limits to publish the number of ads at a time. In this paper, we propose the priority estimation via learning to relative ranking in display ads. Our model is a ranking based model that predicts the ordering the encoded from ad quality scores such as CTR and Cost, which takes different modalities as visual features and other contextual features in ad creative. Our experiments show that the ranking strategy outperforms the regression-based model on a real-world dataset.
Authentication for paper PDF access
A password is required to view paper PDFs. If you are a registered participant, please log on the site from Participant Log In.
You could view the PDF with entering the PDF viewing password bellow.