Participation de l'équipe TGV à DEFT 2022: Prédiction automatique de notes d'étudiants à des questionnaires en fonction du type de question

In this paper we present the work of the TGV team for the DEFT 2022 challenge. We tackled the base task only, which consists of automatically grading students based on their answers to several questions. Our strategy consider this task as a classification task with multiple approaches, each being specific to a question type leading to different types of expected answers. DEFT 2022 allowed us to send three runs for the final test. From our 3 runs, two runs with approaches specialized for each question type: the first one using extracted features and the second one using vectors from TF-IDF or Hashing Vectorizations. The other one is a run using all the questions as one global set and is used to enable comparison with the first two ones. By doing so, our main objective is to verify if approaches dedicated to question types are more suitable for this task than a global one.