Option Tracing and Personalized Question Selection

This is the implementation of our submission in the NeurIPS 2020 Education Challenge.

User: arighosh (Rank 3 in Task 1 and 2, and Rank 1 in Task 4 in the private leaderboard).

Reports

Option Tracing: Beyond Binary Knowledge Tracing

A Meta-learning Framework for Personalized Question Selection

Setups

The requiring environment is as bellow:

• Linux
• Python 3+
• PyTorch 1.5.0
• Scikit-learn 0.22.1
• Scipy 1.4.1
• Numpy 1.18.0

Task 4

For training the model, use the trainer script.

python3 trainer_task_4.py

We added two models in the /model/ folder (including the best performing model in the private leaderboard). For evaluation in the challenge set up, run,

python3 local_evaluation_task_4.py

Task 1 and 2

We added a sample dataset in the data_task_1_2/ folder. You can download the full dataset from here. Download the data_task_1_2.zip in the data_task_1_2/ folder and unzip it. If you want to run with DAS3H features, download the metadata file from here. Download in the data_task_1_2/ folder and unzip it.

For training a model, run,

python3 trainer_task_1_2.py --bidirectional {0,1} --dash {0,1} --model {attn, lstm} --task {'1', '2'}

Contact: Aritra Ghosh (aritraghosh.iem@gmail.com).