Lb_mBERT

Lb_mBERT is a BERT-like language model for the Luxembourgish language.

We used the weights of the multilingual BERT (mBERT) language model as a starting point and continued pre-training it on the MLM task using the same corpus that we used for our LuxemBERT model (https://huggingface.co/lothritz/LuxemBERT).

We achieved higher performances on some downstream tasks than the original LuxemBERT, and another Luxembourgish BERT model called DA BERT (https://huggingface.co/iolariu/DA_BERT).

If you would like to know more about our work, the pre-training corpus, or use our models or datasets, please check out/cite the following papers:

@inproceedings{lothritz-etal-2022-luxembert,
    title = "{L}uxem{BERT}: Simple and Practical Data Augmentation in Language Model Pre-Training for {L}uxembourgish",
    author = "Lothritz, Cedric  and
      Lebichot, Bertrand  and
      Allix, Kevin  and
      Veiber, Lisa  and
      Bissyande, Tegawende  and
      Klein, Jacques  and
      Boytsov, Andrey  and
      Lefebvre, Cl{\'e}ment  and
      Goujon, Anne",
    booktitle = "Proceedings of the Thirteenth Language Resources and Evaluation Conference",
    month = jun,
    year = "2022",
    address = "Marseille, France",
    publisher = "European Language Resources Association",
    url = "https://aclanthology.org/2022.lrec-1.543",
    pages = "5080--5089",
    abstract = "Pre-trained Language Models such as BERT have become ubiquitous in NLP where they have achieved state-of-the-art performance in most NLP tasks. While these models are readily available for English and other widely spoken languages, they remain scarce for low-resource languages such as Luxembourgish. In this paper, we present LuxemBERT, a BERT model for the Luxembourgish language that we create using the following approach: we augment the pre-training dataset by considering text data from a closely related language that we partially translate using a simple and straightforward method. We are then able to produce the LuxemBERT model, which we show to be effective for various NLP tasks: it outperforms a simple baseline built with the available Luxembourgish text data as well the multilingual mBERT model, which is currently the only option for transformer-based language models in Luxembourgish. Furthermore, we present datasets for various downstream NLP tasks that we created for this study and will make available to researchers on request.",
}

@inproceedings{lothritz2023comparing,
  title={Comparing Pre-Training Schemes for Luxembourgish BERT Models},
  author={Lothritz, Cedric and Ezzini, Saad and Purschke, Christoph and Bissyande, Tegawend{\'e} Fran{\c{c}}ois D Assise and Klein, Jacques and Olariu, Isabella and Boytsov, Andrey and Lefebvre, Clement and Goujon, Anne},
  booktitle={Proceedings of the 19th Conference on Natural Language Processing (KONVENS 2023)},
  year={2023}
}