Building Luganda Machine Translation models for the Medical Domain
Keywords: machine translation, luganda, Low-resourced languages, medical machine translation
Abstract: Globalization and migration have highlighted the critical need for effective cross-language communication, particularly in healthcare. In Uganda, a multilingual nation where Luganda is widely spoken, language barriers in predominantly English-speaking medical settings often lead to misunderstandings, misdiagnoses, and compromised patient care. This study aims to mitigate these issues by developing an machine translation model built for medical communication, specifically targeting translations from English to Luganda within the context of malaria diagnosis and community engagement. Utilizing recent advancements in Artificial Intelligence and unsupervised learning, this research involves curating a parallel medical corpus, training a transformer-based model with domain-specific adapters, and rigorously evaluating the model's accuracy and cultural sensitivity.
The results demonstrate that the MarianMT-Adapter LoRa model, when combined with active learning, achieved a significant improvement in translation quality, evidenced by a BLEU score increase to 56. This model effectively reduced translation errors and preserved the contextual integrity of medical texts. The findings are anticipated to enhance healthcare communication, reduce disparities, and improve access to medical knowledge for Luganda-speaking communities, providing a blueprint for similar efforts in other multilingual environments.
Primary Area: other topics in machine learning (i.e., none of the above)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 11515
Loading