This repository contains the source code of the mdoel PEBG in our paper "Improving Knowledge Tracing via Pre-training Question Embeddings", which is accepted by IJCAI 2020.
- data_assist09.py: Data pre-process. You should download original assist09 dataset from here. Details are here.
- ednet/data.py: Data pre-process for EdNet dataset. we use the 'ednet-kt1' dataset. You can download it here. The question information file is here. For more information about EdNet
Once you have the ednet-kt1 dataset, you can enter the folder "ednet" and run 'data.py' to pre-process EdNet dataset.
- extract.py: Extract the implicit similarity between questions and skills.
- PNN.py: Implement the product layer.
- pebg.py: The PEBG model.
- pebg_dkt.py: The PEBG+DKT model.
If you need more information about our experiments, you can contact us. email: liuyunfei@sjtu.edu.cn