Digital Content Profiling Based on User Engagement Features
[ 1 ] Instytut Informatyki, Wydział Informatyki i Telekomunikacji, Politechnika Poznańska | [ P ] employee
2023
chapter in monograph / paper
english
- digital content scoring
- behavioural profiles for articles
- subscription for digital content
- streaming data processing
- big data
EN Exploring audience engagement with digital media content may lead to many various benefits. In this paper, we study how adding engagement-based features to the article description can influence the efficiency of algorithms aimed at detecting digital media readers’ propensity to buy a subscription. Based on the propensity score, the publishers can optimize a decision to display a paywall. Moreover, it is observed that more and more page views are of new or anonymous users. Consequently, the decision concerning the paywall application has to rely only on digital content features. In order to address this application scenario, we propose a novel digital content enrichment framework based on the engagement statistics of users reading a given article. We experimentally evaluate the performance of machine learning algorithms for predicting the propensity to subscribe using the dataset based on events describing the behavior of users exploring the digital news site of the large media publisher. The results of experiments demonstrate that enrichment of article profiles with user engagement features significantly improves prediction models’ efficiency.
20.04.2023
91 - 104
20
70