Automated Essay Scoring System for Nonnative Japanese Learner

To clone this repository together with the required bert-japanese:

git clone --recurse-submodules https://github.com/reo11/aes-for-japanese-learner

Download our pre-trained models

Please download our pre-trained model from here and set a directory name trained_models.

Predicting scores with pre-trained models

You can predict the score by executing the following code.

python predict_with_**.py [input.csv]

When executing LSTM and BERT models, it is better to use GPU as follows.

CUDA_VISIBLE_DEVICES=0 python predict_with_**.py [input.csv]

Input format

The input csv format of essays must be as follows:

text_id	prompt	text
ex1	...	...
ex2	...	...

Output format

The output csv format is as follows:

text_id	holistic	content	organization	language
ex1	3	3	3	3
ex2	4	4	4	4