MT-MCD: A Multi-task Cognitive Diagnosis Framework for Student Assessment

2018 
Student assessment aims to diagnose student latent attributes (e.g., skill proficiency), which is a crucial issue for many educational applications. Existing studies, such as cognitive diagnosis, mainly focus on exploiting students’ scores on questions to mine their attributes from an independent exam. However, in many real-world scenarios, different students usually participate in different exams, where the results obtained from different exams by traditional methods are not comparable to each other. Therefore, the problem of conducting assessments from different exams to obtain precise and comparable results is still underexplored. To this end, in this paper, we propose a Multi Task - Multidimensional Cognitive Diagnosis framework (MT-MCD) for student assessment from different exams simultaneously. In the framework, we first apply a multidimensional cognitive diagnosis model for each independent assessment task. Then, we extract features from the question texts to bridge the connections with each task. After that, we employ a multi-task optimization method for the framework learning. MT-MCD is a general framework where we develop two effective implementations based on two representative cognitive diagnosis models. We conduct extensive experiments on real-world datasets where the experimental results demonstrate that MT-MCD can obtain more precise and comparable assessment results.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    28
    References
    7
    Citations
    NaN
    KQI
    []