In each knowledge tracing task there is a sequence of interactions for each students: $[I_1, \cdots, I_T]$ and each of this interactions consist of several features:
$\{f_t^1, \cdots,  f_t^n\}.$. In knowledge tracing we want to predict feature f at time t, which is formulated as:
$$p(f_t^k | \{ I_1, \cdots, I_{t-1}, I_t, I_{t+1}, \cdots, I_T \}  \div  \Gamma(f_t^k))$$
from which $\Gamma(f)$ is a set of features that should be masked.
In knowledge tracing typically the correctness of the response is predicted. 
The task that is different from knowledge tracing, is assessment predicting. In this case, we want to predict the assessment given the sequence of interactions:
$$p(a_t^k | \{ I_1, \cdots, I_{t-1}, I_t, I_{t+1}, \cdots, I_T \}  \div  \Gamma(a_t^k))$$



 Formal Definition of Assessment Modeling (Assessment Modeling: Fundamental Pre-training Tasks for Interactive Educational Systems)

1. Formal Definition of Assessment Modeling
2. Assessment Modeling as Pre-training Tasks
3. Assessment Modeling with Deep Bidirectional Transformer Encoder

San Diego State University

1. Introduction
2. Related Works
3. Assessment Modeling
4. Experiments

Assessment Modeling: Fundamental Pre-training Tasks for Interactive Educational Systems

The main goal of Interactive Aducational System is to design optimized personalized paths for students. One of the main limitation is the lack of final outcome, similar to final exam grade in the student interactions datasets. In order to get rid of this problem, the authors decided to use Assessment Modeling, which represents the class of pre-training tasks.  Deep Bidirectional Transformer Encoder is used as an assessment model. They empirically evaluate model's performance.

Introduction (Assessment Modeling: Fundamental Pre-training Tasks for Interactive Educational Systems)

1. Artificial Intelligence in Education

Artificial Intelligence in Education (AIEd) tries to improve education using AI and DL. It incorporates knowledge tracing, question analysis, student grade prediction.

2. Pre-training Methods in Education

Pre-training is the technique that is utilized in the various areas: NLP, Computer Vision, Speech Recognition and so on, AIEd is not an exception. In this case pre-training is used to predict student's score on the exams like SAT or whether student will graduate or not. Most of previous pre-training methods are actually NLP tasks where data is produced from learning materials. One of the previous models is Test-aware Attention-based Convolutional Neural Network
(TACNN) which tries to predict the difficulty of the exam question. Additionally, there was another model QuesNet which is embedding model trained on the questions data. This model showed the best performance compared to other models.


Learn Before

Related

Learn After