In this paper, using a specific didactic approach, data-driven learning, we study the use of the FLEURON platform (https://fleuron.atilf.fr/) and, in particular, its concordancer by FLE (French as Foreign Language) teachers and allophone students of different levels. Learners rely on this tool to observe data, identify specific forms, propose categorizations and generalize the functioning of these forms based on real occurrences and sufficiently broad contexts. The concordancer serves as a resource for autonomous learning which, after initial trial and error, results in a real linguistic "investigation". Based on classroom recordings, we propose a multimodal interactional analysis of sequences where the use of certain words in context is illustrated through the mobilization of conversational data within the FLEURON platform. This will lead us to identify the emerging features of the learning process among users of the platform, as for learners' initiatives and their responses to teacher's solicitations.