Skip to content

Latest commit

 

History

History
49 lines (40 loc) · 3.42 KB

README.md

File metadata and controls

49 lines (40 loc) · 3.42 KB
Raw

Transferability of Natural Language Inference to Biomedical Question Answering

This repository provides a code for our paper, which tries to interpret the transferability of natural language inference to biomedical question answering. We use this code for the BioASQ Challenge Task 8b-Phase B. Please refer to our paper Transferability of Natural Language Inference to Biomedical Question Answering for more details. This project is proceeded by DMIS-Lab.

Data Download

We provide a pre-processed version of the BioASQ Task 8b-Phase B

  • Pubmed Abstract : a pre-processed version of pubmed abstract data used for Task 8b-Phase B.
  • SQuAD Oracle : a pre-processed version of SQuAD oracle dataset.
  • Yes/No type : a pre-processed version of Yes/No type questions in Task 8b-Phase B.
  • Factoid type : a pre-processed version of Factoid type questions in Task 8b-Phase B.
  • List type : a pre-processed version of List type questions in Task 8b-Phase B.

We revised the pre-processed datasets from Pre-trained Language Model for Biomedical Question Answering released by BioASQ-BioBERT.

For details of the original BioASQ datasets, please see An overview of the BIOASQ large-scale biomedical semantic indexing and question answering competition (Tsatsaronis et al. 2015).

Pre-trained Model

We use the BioBERT model as our base model learning. For specific fine-tuning procedure, please see our corresponding folder respectively.

In order to facilitate the reproduction of our results, we provide our learned parameters of our model.

  • BioBERT-MNLI : Sequential transfer learning of our model (BioBERT-MNLI) parameters.
  • BioBERT-MNLI-SQuAD : Sequential transfer learning of our model (BioBERT-MNLI-SQuAD) parameters.
  • BioBERT-MNLI-SQuAD(oracle) : Sequential transfer learning of our model (BioBERT-MNLI-SQuAD(oracle)) parameters.

Requirements

  • GPU (Our setting was a single NVIDIA Titan RTX (24GB) GPU)
  • Python version >= 3.7
  • Tensorflow version >= 1.14.0
  • Pytorch version >= 1.5.1

Contact Information

For help or any issues using our code, please contact Minbyul Jeong or Mujeen Sung (minbyuljeong, mujeensung {at} korea.ac.kr). We welcome for any suggestions to modify our issues.

Bibliography

@misc{jeong2020transferability,
    title={Transferability of Natural Language Inference to Biomedical Question Answering},
    author={Minbyul Jeong and Mujeen Sung and Gangwoo Kim and Donghyeon Kim and Wonjin Yoon and Jaehyo Yoo and Jaewoo Kang},
    year={2020},
    eprint={2007.00217},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}