README.md

Aligning open educational resources to new taxonomies: How AI technologies can help and in which scenarios

This repo contains the code for the paper:

Zhi Li, Zachary A Pardos, and Cheng Ren. (In press). Aligning Open Educational Resources to New Taxonomies: How AI technologies can help and in which scenarios. Computers & Education.

Dependencies

The code is written and run with the following packages:

• Python 3.6.8
• PyTorch 1.7.1+cu101
• NumPy 1.18.5
• pandas 1.1.5
• scikit-learn 0.24.2
• sentence-transformers 1.2.1
• efficientnet-pytorch 0.7.0

Data

• data/ contains the Common Core Mathematics skills list and text embeddings of the skill descriptions.
• data/examples contains two example problems for demostration purpose.

Instructions

Data Preparation

Put the problem texts and images in a folder similar to data/examples. Put training labels if any in a csv file with two columns source and target.

Training

Prepare the problems data and labels as described above. Use TextEncoder, ImageEncoder, and Fusion in encoder/ to encode problem data into embeddings, then use ClassificationModelTrainer and SimilarityMatchingModelTrainer in training.py to train the models.

Evaluation

Prepare the problems data as described above. Use TextEncoder, ImageEncoder, and Fusion in encoder/ to encode problem data into embeddings, then use ClassificationModelTester and SimilarityMatchingModelTester in inference.py to run the models.

We also release two pre-trained models that map problems to Common Core skills in models/, and provide a demo.ipynb to show how to use the pre-trained models to predict the skills for the given problems.

References

We referenced the following repos for the code:

• pytorch_compact_bilinear_pooling