HiTZ/mdeberta-expl-extraction-multi

mDeBERTa-base for Multilingual Correct Explanation Extraction in the Medical Domain

This model is a fine-tuned version of mdeberta-v3-base for a novel extractive task which consists of identifying the explanation of the correct answer written by medical doctors. The model has been fine-tuned using the multilingual https://huggingface.co/datasets/HiTZ/casimedicos-squad dataset, which includes English, French, Italian and Spanish.

Performance

The model scores 74.64 F1 partial match (as defined in SQuAD extractive QA task) averaged across the 4 languages.

• tags: to delimit explanations of the correct answers and the rest at token level.
  • 0: explanation of the correct answer
  • 1: others

Fine-tuning hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 48
• eval_batch_size: 8
• seed: random
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20.0

Framework versions

• Transformers 4.30.0.dev0
• Pytorch 2.1.2+cu121
• Datasets 2.16.1
• Tokenizers 0.15.2

Citation

If you use this model please cite the following paper:

@misc{goenaga2023explanatory,
      title={Explanatory Argument Extraction of Correct Answers in Resident Medical Exams}, 
      author={Iakes Goenaga and Aitziber Atutxa and Koldo Gojenola and Maite Oronoz and Rodrigo Agerri},
      year={2023},
      eprint={2312.00567},
      archivePrefix={arXiv}
}

Contact: Iakes Goenaga and Rodrigo Agerri HiTZ Center - Ixa, University of the Basque Country UPV/EHU