HiTZ/mdeberta-metaphor-detection-multilang-es_en

mDeBERTa for Multilingual Metaphor Detection (es/en)

This model is a fine-tuned version of mDeBERTa on Meta4XNLI dataset, for metaphor detection in Spanish and English at token level. This model is presented in our paper Meta4XNLI: A Crosslingual Parallel Corpus for Metaphor Detection and Interpretation

Model Sources

• Repository: (https://github.com/elisanchez-beep/meta4xnli)
• Paper: Meta4XNLI: A Crosslingual Parallel Corpus for Metaphor Detection and Interpretation .

Training & Testing Data

Meta4XNLI

Training Hyperparameters

• Batch size: 8
• Weight Decay: 0.1
• Learning Rate: 0.00005
• Epochs: 4

Results

• F1_es: 68.70
• F1_en: 60.80
• Precision_es: 72.78
• Precision_en: 65.03
• Recall_es: 65.29
• Recall_en: 57.23

Label Dictionary

{
  "LABEL_0": "B-METAPHOR",
  "LABEL_1": "I-METAPHOR",
  "LABEL_2": "O"
}

Citation

If you use this model, please cite our work:

@misc{sanchezbayona2024meta4xnlicrosslingualparallelcorpus,
      title={Meta4XNLI: A Crosslingual Parallel Corpus for Metaphor Detection and Interpretation}, 
      author={Elisa Sanchez-Bayona and Rodrigo Agerri},
      year={2024},
      eprint={2404.07053},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2404.07053}
}

Dataset Card Contact

{elisa.sanchez, rodrigo.agerri}@ehu.eus