@inproceedings{akinade-etal-2023-varepsilon,
    title = "Varepsilon k{\'u} mask: Integrating {Y}or{\`u}b{\'a} cultural greetings into machine translation",
    author = "Akinade, Idris  and
      Alabi, Jesujoba  and
      Adelani, David  and
      Odoje, Clement  and
      Klakow, Dietrich",
    editor = "Dev, Sunipa  and
      Prabhakaran, Vinodkumar  and
      Adelani, David  and
      Hovy, Dirk  and
      Benotti, Luciana",
    booktitle = "Proceedings of the First Workshop on Cross-Cultural Considerations in NLP (C3NLP)",
    month = may,
    year = "2023",
    address = "Dubrovnik, Croatia",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.c3nlp-1.1",
    doi = "10.18653/v1/2023.c3nlp-1.1",
    pages = "1--7",
    abstract = "This paper investigates the performance of massively multilingual neural machine translation (NMT) systems in translating Yor{\`u}b{\'a} greetings (k{\'u} mask), which are a big part of Yor{\`u}b{\'a} language and culture, into English. To evaluate these models, we present IkiniYor{\`u}b{\'a}, a Yor{\`u}b{\'a}-English translation dataset containing some Yor{\`u}b{\'a} greetings, and sample use cases. We analysed the performance of different multilingual NMT systems including Google and NLLB and show that these models struggle to accurately translate Yor{\`u}b{\'a} greetings into English. In addition, we trained a Yor{\`u}b{\'a}-English model by fine-tuning an existing NMT model on the training split of IkiniYor{\`u}b{\'a} and this achieved better performance when compared to the pre-trained multilingual NMT models, although they were trained on a large volume of data.",
}