HiTZ/mbert-argmining-abstrct-multilingual

mBERT for multilingual Argument Detection in the Medical Domain

This model is a fine-tuned version of bert-base-multilingual-cased for the argument component detection task on AbstRCT data in English, Spanish, French and Italian (https://huggingface.co/datasets/HiTZ/multilingual-abstrct).

Performance

F1-macro scores (at sequence level) and their averages per test set from the argument component detection results of monolingual, monolingual automatically post-processed, multilingual, multilingual automatically post-processed, and crosslingual experiments.

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3.0

Framework versions

• Transformers 4.40.0.dev0
• Pytorch 2.1.2+cu121
• Datasets 2.16.1
• Tokenizers 0.15.2

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