Abstract
This work addresses the task of identifying English code-switching in multilingual text. We train two token-level classifiers on data of high-resource language pairs. The first distinguishes between English, not English, morphologically mixed, and other words. The second is a binary classifier that identifies named entities. Results indicate that our system is on-par with SoTA for high-resource language pairs. Meanwhile we show that on low-resource language pairs not in the training data our system outperforms SoTA by between 2.31 and 4.59% F1. We also analyse the correlation between typological similarity of the languages and difficulty in recognizing code-switching. Our system is a new strong baseline system for code-switching research between any language and English.- Anthology ID:
- 2024.vardial-1.14
- Volume:
- Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024)
- Month:
- June
- Year:
- 2024
- Address:
- Mexico City, Mexico
- Editors:
- Yves Scherrer, Tommi Jauhiainen, Nikola Ljubešić, Marcos Zampieri, Preslav Nakov, Jörg Tiedemann
- Venues:
- VarDial | WS
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 163–173
- Language:
- URL:
- https://aclanthology.org/2024.vardial-1.14
- DOI:
- Cite (ACL):
- Igor Sterner. 2024. Multilingual Identification of English Code-Switching. In Proceedings of the Eleventh Workshop on NLP for Similar Languages, Varieties, and Dialects (VarDial 2024), pages 163–173, Mexico City, Mexico. Association for Computational Linguistics.
- Cite (Informal):
- Multilingual Identification of English Code-Switching (Sterner, VarDial-WS 2024)
- PDF:
- https://preview.aclanthology.org/jeptaln-2024-ingestion/2024.vardial-1.14.pdf