Abstract. In the context of the French presidential elections held in 2017, we developed the Politoscope’, an open platform for the analysis of political tweets, which was used to collect and analyse almost 60 million Twitter exchanges between more than 2.4 million users who interacted with French political figures. In this paper, we focus on the information revealed by this data, in terms of the development of the political environment during the elections. Among the numerous types of interaction that can be used with Twitter, we focus on that of retweeting. We characterise the socio-semantic networks of the French political environment, and their development over a period of eleven months preceding the presidential election. We show that analysis of retweet graphs can allow accurate and automatic monitoring of the reconfigurations of a multipolar political environment, by characterising the different political communities, their alliance strategies and their semantic particularities. This is the first study of the socio-semantic trends of a multipolar political environment, and contributes to the emergence of new methods, originating from computational social sciences, which significantly modify research practices in the social sciences, and in political science in particular.