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.- Anthology ID:
- 2024.iwclul-1.13
- Volume:
- Proceedings of the 9th International Workshop on Computational Linguistics for Uralic Languages
- Month:
- November
- Year:
- 2024
- Address:
- Helsinki, Finland
- Editors:
- Mika Hämäläinen, Flammie Pirinen, Melany Macias, Mario Crespo Avila
- Venue:
- IWCLUL
- SIG:
- SIGUR
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 104–108
- Language:
- URL:
- https://preview.aclanthology.org/add_missing_videos/2024.iwclul-1.13/
- DOI:
- Cite (ACL):
- Aleksei Dorkin, Taido Purason, and Kairit Sirts. 2024. Prune or Retrain: Optimizing the Vocabulary of Multilingual Models for Estonian. In Proceedings of the 9th International Workshop on Computational Linguistics for Uralic Languages, pages 104–108, Helsinki, Finland. Association for Computational Linguistics.
- Cite (Informal):
- Prune or Retrain: Optimizing the Vocabulary of Multilingual Models for Estonian (Dorkin et al., IWCLUL 2024)
- PDF:
- https://preview.aclanthology.org/add_missing_videos/2024.iwclul-1.13.pdf