Busayo Awobade, Mardhiyah Sanni, Tassallah Abdullahi, Chibuzor Okocha, Kelechi Ezema, Devendra Kayande, Lukman Enegi Ismaila, Gloria Ashiya Katuka, Tobi Olatunji
EACL 2026
Introduction of AfriVox, an open-source benchmark for over 20 African languages and 100+ accents, used to evaluate leading multimodal LLMs, reveal significant performance disparities,
and achieve state-of-the-art results through targeted fine-tuning.
Gabrial Zencha Ashungafac, Mardhiyah Sanni, Busayo Awobade, Alex Gichamba, Tobi Olatunji
IJCNLP-AACL 2025
Development and empirical analysis of Afrispeech-MultiBench, the first domain-specific evaluation benchmark for assessing speech recognition performance
across 100+ African English accents and six application domains.
Chris Emezue, NaijaVoices Community, Busayo Awobade, Abraham Owodunni, Handel Emezue, Gloria Emezue, Nefertiti Emezue, Sewade Ogun, Bunmi Akinremi, David Adelani, Chris Pal
Interspeech 2025
Creation and validation of NaijaVoices, a novel 1,800-hour speech dataset for low-resource Nigerian languages
(Igbo, Hausa, and Yoruba), addressing critical data scarcity and enabling substantial performance gains in automatic speech recognition models.
Busayo Awobade*, Mardiyyah Oduwole*, Steven Kolawole*
arXiv
Investigates how data augmentation techniques can enhance machine translation for low-resource African languages.
Using sentence concatenation with back translation and switch-out across six languages,
the approach demonstrates strong potential for advancing translation in under-resourced settings.
Busayo Awobade*, Mardiyyah Oduwole*, Steven Kolawole*
AfricaNLP 2024
Demonstrates that compression techniques significantly improve the efficiency and effectiveness of small,
low-resource language models, confirming that their benefits are not limited to large-scale models.
* denotes equal contribution
This list includes peer-reviewed publications, workshop papers, and preprints.
For the most up-to-date list with citation counts, please visit my Google Scholar profile.