README.md

improved_keras_english_resume_IE

This code is the paper with:
@article{甘程光2021英文履歴書データ抽出システムへの,
title={英文履歴書データ抽出システムへの BERT 適用性の検討},
author={甘程光 and 高橋良英 and others},
journal={2021 年度 情報処理学会関西支部 支部大会 講演論文集},
volume={2021},
year={2021}
}

The dataset is already publicly available at kaggle. You can download it at this URL below.
https://www.kaggle.com/datasets/chingkuangkam/resume-text-classification-dataset
and HuggingFace https://huggingface.co/datasets/ganchengguang/resume-5label-classification

Put the dataset into this folder (demo\data\training_data) and you can use it.

Bi-LSTM-Attention and CNN-Attention Model:

The main code are base in chen0040/keras-english-resume-parser-and-analyzer for improvement.
Add the Attention layer in last original output layer.
And you need pip install attention

BERT Model:

BERT Model is base in sujianlin's bert4keras.

You need pip install bert4keras

Fine-tune use bert train.py
predict a new resume text use bert predict.py