Towards an Appropriate Query, Key, and Value Computation for Knowledge Tracing

Knowledge Tracing(KT) comes in various forms, from data-driven algorithms to deep learning-based models. This study integrates a contemporary deep learning model called Transformers and applies it to student response data.

Definition:

Knowledge Tracing models attempt to accurately determine a student’s current knowledge state, such as whether a student may respond correctly to a question

Separated Self-AttentIve Neural Knowledge Tracing(SAINT) utilizes an encoder-decoder structure that enables response sequences to stack attention layers multiple times, capturing the complex relations between educational exercises and student responses. Empirical evaluations show that SAINT attained state-of-the-art performance in our EdNet dataset.