In this article, we propose a method for the automatic correction of definitions produced by students in the field of management and economics. Our algorithm processes the students' textual responses to identify which concepts and relations are correctly restituted and which are erroneous or missing. We carried out an evaluation on several corpora of definitions, and the results show an F1 score of around 0.94 for the identification of relations, and between 0.75 and 0.88 for the identification of nominal groups which are elements of the relations. This approach is part of a wider project involving the automatic correction of textual responses and the production of tools for teachers and learners.