An NLP framework to automatically evaluate the adequacy and relevance of assessment items

Christophe Bobineau
Calendar
When:
10/01/2026 – 11/01/2026 all-day
2026-01-10T01:00:00+01:00
2026-01-11T01:00:00+01:00
Stages

Offre en lien avec l’Action/le Réseau : – — –/– — –

Laboratoire/Entreprise : SAMOVAR, Télécom SudParis, IP Paris
Durée : 6 Months
Contact : luca.benedetto@telecom-sudparis.eu
Date limite de publication : 2026-01-10

Contexte :

Sujet :
This project proposes developing an NLP framework to automatically evaluate the adequacy and relevance of assessment items in relation to their associated learning content. While existing research in Question Difficulty Estimation from Text (QDET) has focused primarily on analyzing exam items in isolation, this work addresses a critical gap by evaluating questions within the context of course lectures and learning paths. The framework will employ a combination of traditional machine learning, Information Retrieval techniques, semantic embeddings, and Large Language Models to assess newly created exam questions for validity, relevance, and difficulty.

Profil du candidat :
Previous experience with Python and Machine Learning is required.

Formation et compétences requises :

Adresse d’emploi :
19 place Marguerite Perey, 91120 Palaiseau France

ou

9 rue Charles Fourier, 91011 Evry-Courcouronnes France

Document attaché : 202512031306_2025_11___Proposal_Stage_M2.pdf