Vietnamese Relation Extraction with BERT-based Models at VLSP 2020
This project contains source code for our approach to solve Relation Extraction task at VSLP 2020. The detail is described in the following paper:
Please contact VLSP 2020 Organizers for the dataset. You can follow their guide at: https://vlsp.org.vn/resources
In the project, we have five notebooks, you can upload and use it with Google Colab.
- VLSP2020_RE_extract_[training, dev, test]: You can use these notebooks to extract information from the raw datasets. Please remember to upload the original VLSP 2020 dataset to notebooks respectively. You don't need GPU to run these notebook.
- VSLP2020_RE_model_training: This notebook uses the output extracted from the raw training and development data to train the model. Our model code is placed in "Model" part.
- VLSP2020_RE_model_test: This notebook uses saved flags.txt and pretrained model from the training notebook's output to get the final results on test data. However, due to an accident, I lost the code that helps to save the final results to files in the format required by the organizers.
Please feel free to contact me at: nguyenhuuthuat09@gmail.com
@inproceedings{nguyen-man-2020-vietnamese,
title = "{V}ietnamese Relation Extraction with {BERT}-based Models at {VLSP} 2020",
author = "Nguyễn, Thuật and
Mẫn, Hiếu",
booktitle = "Proceedings of the 7th International Workshop on Vietnamese Language and Speech Processing",
month = dec,
year = "2020",
address = "Hanoi, Vietnam",
publisher = "Association for Computational Lingustics",
url = "https://www.aclweb.org/anthology/2020.vlsp-1.6",
pages = "30--34",
}