@inproceedings{dorkin-etal-2024-erzya,
    title = "Prune or Retrain: Optimizing the Vocabulary of Multilingual Models for {E}stonian",
    author = "Dorkin, Aleksei  and
      Purason, Taido  and
      Sirts, Kairit",
    editor = {H{\"a}m{\"a}l{\"a}inen, Mika  and
      Pirinen, Flammie  and
      Macias, Melany  and
      Crespo Avila, Mario},
    booktitle = "Proceedings of the 9th International Workshop on Computational Linguistics for Uralic Languages",
    month = nov,
    year = "2024",
    address = "Helsinki, Finland",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/jlcl-multiple-ingestion/2024.iwclul-1.13/",
    pages = "104--108",
    abstract = "Adapting multilingual language models to specific languages can enhance both their efficiency and performance. In this study, we explore how modifying the vocabulary of a multilingual encoder model to better suit the Estonian language affects its downstream performance on the Named Entity Recognition (NER) task. The motivations for adjusting the vocabulary are twofold: practical benefits affecting the computational cost, such as reducing the input sequence length and the model size, and performance enhancements by tailoring the vocabulary to the particular language. We evaluate the effectiveness of two vocabulary adaptation approaches{---}retraining the tokenizer and pruning unused tokens{---}and assess their impact on the model`s performance, particularly after continual training. While retraining the tokenizer degraded the performance of the NER task, suggesting that longer embedding tuning might be needed, we observed no negative effects on pruning."
}