Most of the link analysis methods developped for the Quality Assessment task consider only one set of nodes and a single relation. However, recent crowdsourcing platforms lay on various different semantics of nodes and relations. In this work, we propose a model of crowdsourcing platforms using heterogeneous graphs. Based on this representation, we propose an algorithm that can easily take advantage of various semantics of nodes and relations and make it adaptable to the evolution of the platforms. We show that our proposition generalizes some state-of-the-art models. Furthemore, experiments conducted on the two platforms, Wikipedia, and Stack Exchange, show a real interest to consider user interactions in this task.