tthhanh/xlm-ate-nobi-fr-nes

XLMR Token Classifier for Term Extraction

This model is a fine-tuned version of xlm-roberta-base for cross-domain term extraction tasks.

Model description

This model is a fine-tuned version of xlm-roberta-base for token classification, specifically designed to identify and classify terms within text sequences. The model assigns labels such as B-Term, I-Term, BN-Term, IN-Term, and O to individual tokens, allowing for the extraction of meaningful terms from the text.

Intended uses & limitations

The model is intended for term extraction tasks. It can be applied in domains like:

• Named Entity Recognition (NER)
• Information Extraction

Training and evaluation data

We fine-tuned the French version of the ACTER dataset where Named Entities are included in the gold standard. We trained on the Corruption and Wind Energy domain, validated on the Equitation domain, and tested on the Heart Failure domain.

Training procedure

The following hyperparameters were used during training:

- learning_rate: 2e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 20

Framework versions:

- Transformers 4.26.1
- Pytorch 2.0.1+cu117
- Datasets 2.9.0
- Tokenizers 0.13.2

Evaluation

We evaluate the performance of the ATE systems by comparing the candidate list extracted from the test set with the manually annotated gold standard term list for that specific test set. We use exact string matching to compare the retrieved terms to the ones in the gold standard and calculate Precision (P), Recall (R), and F1-score (F1). The results are reported in Can cross-domain term extraction benefit from cross-lingual transfer and nested term labeling?.

Citation

If you use this model in your research or application, please cite it as follows:

@inproceedings{tran2022can,
  title={Can cross-domain term extraction benefit from cross-lingual transfer?},
  author={Tran, Hanh Thi Hong and Martinc, Matej and Doucet, Antoine and Pollak, Senja},
  booktitle={International Conference on Discovery Science},
  pages={363--378},
  year={2022},
  organization={Springer}
}