This directory contains source code for the following paper:
Conceptual Grounding Constraints for Truly Robust Biomedical Name Representations.
Pieter Fivez, Simon Šuster and Walter Daelemans. EACL, 2021.
If you use this code, please cite:
@inproceedings{fivez-etal-2021-conceptual,
title = "Conceptual Grounding Constraints for Truly Robust Biomedical Name Representations",
author = "Fivez, Pieter and
Suster, Simon and
Daelemans, Walter",
booktitle = "Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume",
year = "2021",
publisher = "Association for Computational Linguistics",
pages = "2440--2450"}
GPL-3.0
All requirements are listed in requirements.txt.
You can run pip install -r requirements.txt
, preferably in a virtual environment.
The fastText model used in the paper can be downloaded from the following link:
https://drive.google.com/file/d/1B07lc3eeW_zughHguugLBR4iJYQj_Wxz/view?usp=sharing
Our example scripts require a path to this downloaded model.
For convenience, we only provide our adaptation of the openly available MedMentions corpus.
The source files for this corpus can be found at https://github.com/chanzuckerberg/MedMentions.
The script data/extract_medmentions.py has used these source files to create data/medmentions.json.
We provide 2 scripts to run experiments from the paper.
main_dan.py trains and evaluates the DAN encoder on data/medmentions.json.
Please run python main_dan.py --help
to see the options, or check the script.
The default parameters are the best parameters reported in our paper.
main_bne.py trains and evaluates our counterpart implementation of the BNE encoder, as described in:
@inproceedings{phan-etal-2019-robust,
title = "Robust Representation Learning of Biomedical Names",
author = "Phan, Minh C. and
Sun, Aixin and
Tay, Yi",
booktitle = "Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics",
year = "2019",
publisher = "Association for Computational Linguistics",
pages = "3275--3285"}
Please run python main_bne.py --help
to see the options, or check the script.
The default parameters are the best parameters reported in our paper.