LitBERT-CRF

LitBERT-CRF model is a fine-tuned BERT-CRF architecture specifically designed for Named Entity Recognition (NER) in Portuguese-written literature.

Model Details

Model Description

LitBERT-CRF leverages a BERT-CRF architecture, initially pre-trained on the brWaC corpus and fine-tuned on the HAREM dataset for enhanced NER performance in Portuguese. It incorporates domain-specific literary data through Masked Language Modeling (MLM), making it well-suited for identifying named entities in literary texts.

• Model type: BERT-CRF for NER
• Language: Portuguese
• Fine-tuned from model: BERT-CRF on brWaC and HAREM

Evaluation

Testing Data, Factors & Metrics

Testing Data

PPORTAL_ner dataset

Metrics

• Precision: 0.783
• Recall: 0.774
• F1-score: 0.779

Citation

BibTeX:

@inproceedings{silva-moro-2024-evaluating,
    title = "Evaluating Pre-training Strategies for Literary Named Entity Recognition in {P}ortuguese",
    author = "Silva, Mariana O.  and
      Moro, Mirella M.",
    editor = "Gamallo, Pablo  and
      Claro, Daniela  and
      Teixeira, Ant{\'o}nio  and
      Real, Livy  and
      Garcia, Marcos  and
      Oliveira, Hugo Gon{\c{c}}alo  and
      Amaro, Raquel",
    booktitle = "Proceedings of the 16th International Conference on Computational Processing of Portuguese - Vol. 1",
    month = mar,
    year = "2024",
    address = "Santiago de Compostela, Galicia/Spain",
    publisher = "Association for Computational Lingustics",
    url = "https://aclanthology.org/2024.propor-1.39",
    pages = "384--393",
}

APA:

Mariana O. Silva and Mirella M. Moro. 2024. Evaluating Pre-training Strategies for Literary Named Entity Recognition in Portuguese. In Proceedings of the 16th International Conference on Computational Processing of Portuguese - Vol. 1, pages 384–393, Santiago de Compostela, Galicia/Spain. Association for Computational Lingustics.