Comment former des groupes d’étudiants homogènes à partir des résultats de SELF ? Présentation d’un outil d’aide à la décision pour la création de groupes

Vincent Brault, Sylvain Coulange, Frédérique Letué, Marie-Pierre Jouannaud, Marie-José Martinez, Anne-Cécile Perret, Université Grenoble Alpes

Abstract: Faced with the diversity of foreign language skills and heterogeneous language profiles of students entering higher education, universities need an effective tool to place their students into relatively homogeneous groups. In this context, an online placement test with formative orientation called SELF (Système d'Évaluation en Langues à visée Formative) was developed at Université Grenoble Alpes between 2012 and 2020. This language placement test aims to evaluate three skills fully automatically in one hour: listening comprehension, reading comprehension and short written expression, in one of the six languages currently offered. At the end of the test, students receive a placement recommendation (the level of the group they should enroll in), as well as their CEFR level in the three skills tested. The research presented in this article aims to go beyond these skill levels, and attempt to profile students based on their response patterns by comparing SELF placement results with the result of a co-clustering algorithm. Experimental software was developed to allow the user to upload a SELF data file (detailing each response for each student) and to obtain the optimal classification of students and items, or generate classifications given the number of groups desired. Co-classification is still only effective on the initial SELF testlet, i.e. the questions answered by all candidates, and the team is currently working on the integration of all test results.