Dealing with Item Cold-Start in News Recommender at Globo
Globo is the largest Latin American mass media group, where its vertical information portals play an important role in content distribution. Among such portals, G1 is Globo’s journalism portal, being the most popular news portal in Brazil and responsible for delivering informative content to more than 100 million unique users per day. In this context, recommender systems play an important role in achieving a good user experience, offering personalized content. In this paper, we discuss how G1’s recommender system identifies and deals with the item cold-start problem, describing the recommendation scenarios and how the applied improvements in the currently deployed algorithms led to a decreased processing time and an increased CTR in the context of news recommendations.
Jorge Díez Peláez, David Martínez Rego, Amparo Alonso-Betanzos, Óscar Luaces Rodríguez, Antonio Bahamonde Rionda, et al. 2016. Metrical representation of readers and articles in a digital newspaper. In 10th ACM Conference on Recommender Systems (RecSys 2016). ACM.
Pasquale Lops, Marco de Gemmis, and Giovanni Semeraro. 2011. Content-based Recommender Systems: State of the Art and Trends. Springer US, Boston, MA, 73–105. https://doi.org/10.1007/978-0-387-85820-3_3
Badrul Sarwar, George Karypis, Joseph Konstan, and John Riedl. 2000. Analysis of recommendation algorithms for e-commerce. In Proceedings of the 2nd ACM Conference on Electronic Commerce. 158–167.
Badrul Sarwar, George Karypis, Joseph Konstan, and John Riedl. 2001. Item-based collaborative filtering recommendation algorithms. In Proceedings of the 10th international conference on World Wide Web. 285-295.
Andrew I Schein, Alexandrin Popescul, Lyle H Ungar, and David M Pennock. 2002. Methods and metrics for cold-start recommendations. In Proceedings of the 25th annual international ACM SIGIR conference on Research and development in information retrieval. 253–260.