Deep Factorization Machines for Knowledge Tracing

TL;DR

This paper presents a deep factorization machine model for knowledge tracing in language learning, demonstrating improved performance over logistic regression but not surpassing the top models, and offering insights for future enhancements.

Contribution

Introduces a deep factorization machine approach for knowledge tracing, combining wide and deep learning to model pairwise relationships in language learning data.

Findings

Achieved an AUC of 0.815, outperforming logistic regression.

Identified strategies to enhance item response theory models.

Provided insights into modeling user-item interactions in language acquisition.

Abstract

This paper introduces our solution to the 2018 Duolingo Shared Task on Second Language Acquisition Modeling (SLAM). We used deep factorization machines, a wide and deep learning model of pairwise relationships between users, items, skills, and other entities considered. Our solution (AUC 0.815) hopefully managed to beat the logistic regression baseline (AUC 0.774) but not the top performing model (AUC 0.861) and reveals interesting strategies to build upon item response theory models.