The goal of knowledge tracing is to model students' mastering levels of underlying knowledge concepts, termed knowledge state, based on students' exercise performance data. However, existing methods, such as Bayesian Knowledge Tracing (BKT) or Deep Knowledge Tracing (DKT), either require costly human-labeled concept annotations or fail to exactly pinpoint which concepts a student is good at or unfamiliar with. To solve these problems, in this paper we introduce a new model called Dynamic Key-Value Memory Network (DKVMN) that can learn representations using nonlinear transformations and directly output a student's mastering level of each concept. Unlike standard Memory-Augmented Neural Networks (MANNs) that facilitate a single memory matrix or two static memory matrices, our model has one static matrix called key that stores the knowledge concepts and the other dynamic matrix called value that stores and updates corresponding concepts' mastery levels. Experiments show that our DKVMN model, which is trained end-to-end, consistently outperforms the state-of-the-art model on a range of knowledge tracing data-sets. We also illustrate that the learned DKVMN can automatically discover underlying concepts of the exercises which are typically performed by human annotations, and depict a student's changing knowledge state.
from cs.AI updates on arXiv.org http://ift.tt/2gwsQZ7
via IFTTT
No comments:
Post a Comment