becauselol / multi-scale-bert-aes Goto Github PK

This is a demo for the paper(NAACL 2022 long paper accepted): "On the Use of BERT for Automated Essay Scoring: Joint Learning of Multi-Scale Essay Representation"

Steps to run the decoding process:
step 1: Download the project to the computer.
step 2: Download the model(for ASAP's prompt 8) to the computer with the following link:
        Link: https://pan.baidu.com/s/1_m_-DQlX-dLh1XdhOMzj1A?pwd=tmmb
step 3: Update the config file "asap.ini" and set the parameters as following:
        data_dir: your local directory to store the essays of prompt 8 (file format: id \t text \t score);
        model_directory: your local directory to store the model which is obtained in step 2;
        result_file: the path of file to store prediction result.
step 4: Run the following script to get scoring result:
        sh decode.sh