RKT: Relation-Aware Self-Attention for Knowledge Tracing

Abstract

The world has transitioned into a new phase of online learning in response to the recent Covid19 pandemic. Now more than ever, it has become paramount to push the limits of online learning in every manner to keep flourishing the education system. One crucial component of online learning is Knowledge Tracing (KT). The aim of KT is to model student's knowledge level based on their answers to a sequence of exercises referred as interactions. Students acquire their skills while solving exercises and each such interaction has a distinct impact on student ability to solve a future exercise. This impact is characterized by 1) the relation between exercises involved in the interactions and 2) student forget behavior. Traditional studies on knowledge tracing do not explicitly model both the components jointly to estimate the impact of these interactions. In this paper, we propose a novel Relation-aware self-attention model for Knowledge Tracing (RKT). We introduce a relation-aware self-attention layer that incorporates the contextual information. This contextual information integrates both the exercise relation information through their textual content as well as student performance data and the forget behavior information through modeling an exponentially decaying kernel function. Extensive experiments on three real-world datasets, among which two new collections are released to the public, show that our model outperforms state-ofthe-art knowledge tracing methods. Furthermore, the interpretable attention weights help visualize the relation between interactions and temporal patterns in the human learning process.

EKT: Exercise-Aware Knowledge Tracing for Student Performance Prediction

“…Along this line, the highorder graph structure information and knowledge concept information can be fully explored for the final student performance prediction. Educational psychology models are mainly discussed from two sides: cognitive diagnosis models and knowledge tracing models [17,20]. Cognitive diagnosis models, assuming students' knowledge states are static throughout their practice, aim to discover students' proficiency to predict their future performance [18,5,28,35,10,8].…”

Knowledge structure enhanced graph representation learning model for attentive knowledge tracing

“…38 To focus on the relevant interactions, the Transformer framework applies a self-attention mechanism to the input data, and hence incorporates the inner relations in the exercise sequences into the network. Attention based KT models inspired by this study have become an active research area, 15,16,41 a representative work is the attentive KT (AKT) model, 16 which obtained the state-of-the-art performance on this task. However, in most of these models, the KT is based on the skills in a specific domain.…”

“…This approach loses the distinctive information related to individual questions, leading to imprecise inferences of the learners' knowledge states. 12,[14][15][16] For example, in Figure 1(left), the questions "3+5" and "345+6789" both require the skill "addition of integers," and are considered as the same inputs when building the KT models, which ignores their different difficulty levels. Existing research have proved that question difficulty undoubtedly influences the learner performance, 11,17,18 and the relative difficulty level of a specific question varies from learner to learner.…”

Integrating Performance and Side Factors into Embeddings for Deep Learning-Based Knowledge Tracing

“…They use the embedding of skill-response-hint instead of skill-response as the input to acquire the learner's knowledge growth after a response. Pandey and Srivastava propose a Relation-aware self-attention Knowledge Tracing (RKT) model [12] based on SAKT. They take into account the relations between skills and time elapsed since the last interaction to inform the self-attention mechanism.…”