Authors: Bráulio Miranda Veloso
Tags: 2019, conceptual modeling
Several recommender systems have been proposed in the literature. Some of them address the problem of recommending news to users of newspaper sites. The context of online news presents some particularities: highly dynamic data volume, users access without registration, quick accesses of a few readings, and content is time-dependent. Online newspapers generate in real-time the recommendation lists with few items. The recommenders have a short time to model access, create the list, and present it to the user. All these characteristics make the problem of news recommendation an exciting challenge that has been studied by the academic community with new proposals. However, scarce works study users reading behavior before proposing new methods. In this work, we are interested in characterizing online newspaper users via stochastic models and using attributes extracted from this characterization in recommender systems. First results demonstrate that models who use only information from the recent past are the best. Next, we will look at whether these models are best, varying data contexts, and how to generate more personalized models. Finally, we intend to add all the knowledge in a recommender system, improving or creating a new one.Read the full paper here: https://link-springer-com.proxy2.hec.ca/chapter/10.1007/978-3-030-34146-6_24
