Centrality metrics in graphs, such as degree, help to understand the influence of entities in various applications. They are used to quantify the influence of entities based on their relationships. Time dimension has been integrated to take into account the evolution of relationships between entities in real-world phenomena. For instance, in the context of disease spreading, new contacts may appear and disappear over time between individuals. However, they do not take into account the semantics of entities and their relationships. For example, in the context of a disease spreading, some relationships (such as physical contacts) may be more important than others (such as virtual contacts). To overcome this drawback, we propose centrality metrics that integrate both temporal and semantics aspects. We carry out experimental assessments, with real-world datasets, to illustrate the efficiency of our solution.