A Hybrid mT5-Based Machine Translation System for Kanuri–English Educational Translation in a Low-Resource Setting

Kanuri is a morphologically rich, low-resource language spoken primarily in northeastern Nigeria and the Lake Chad basin, yet it remains almost entirely absent from modern natural language processing (NLP) and machine translation (MT) research. This paper presents a hybrid AI-based Kanuri–English machine translation system designed for primary school educational use. The proposed system integrates the multilingual Text-to-Text Transfer Transformer (mT5) with a deterministic dictionary-based lookup module to address the critical challenge of data scarcity inherent to low-resource language translation. A domain-specific parallel corpus of 522 bilingual entries—comprising classroom instructions, greetings, and basic educational vocabulary—was manually compiled and used to fine-tune the mT5 model, while a structured Kanuri–English lexicon was developed to supplement neural translation outputs. The hybrid architecture exploits the complementary strengths of learned neural representations and rule-based lexical mappings, substantially improving translation accuracy and semantic reliability within the educational domain. An interactive web-based interface incorporating text input, browser-based speech recognition, automatic translation, and text-to-speech output was implemented to support multimodal engagement for young learners and educators. Evaluation using accuracy, precision, recall, and F1-score demonstrates that the hybrid system achieves 100% accuracy on the in-domain evaluation set, compared with 34.67% for the standalone mT5 model. Error analysis confirms that hybrid integration mitigates the generalization weaknesses of purely neural approaches in low-resource conditions. The findings provide a replicable framework for inclusive NLP tool development for other underrepresented African languages and contribute to the broader goals of educational equity and linguistic inclusion for Kanuri-speaking primary school students.

Kanuri language, low-resource machine translation, mT5, hybrid translation, multilingual NLP, African language technology, educational NLP, transformer fine-tuning.

1. D. Ataman, A. Birch, N. Habash, M. Federico, P. Koehn, and K. Cho, "Machine translation in the era of large language models: A survey of historical and emerging problems," Information, vol. 16, no. 9, p. 723, 2025, doi: 10.3390/info16090723. [Google Scholar] [Crossref]

2. P. Koehn, Neural Machine Translation. Cambridge, UK: Cambridge Univ. Press, 2020. [Google Scholar] [Crossref]

3. P. Joshi, S. Santy, A. Budhiraja, K. Bali, and M. Choudhury, "The state and fate of linguistic diversity and inclusion in the NLP world," arXiv:2004.09095, 2020. [Google Scholar] [Crossref]

4. I. Inuwa-Dutse, "NaijaNLP: A survey of Nigerian low-resource languages," arXiv:2502.19784, 2025. [Google Scholar] [Crossref]

5. B. M. Saleh, S. Bilgaiyan, and S. Sagnika, "KanuriSenti: A novel dataset for sentiment analysis in the under-resourced Kanuri language," Data Brief, vol. 61, Art. no. 111758, 2025, doi: 10.1016/j.dib.2025.111758. [Google Scholar] [Crossref]

6. UNESCO, Reimagining Our Futures Together: A New Social Contract for Education. Paris, France: UNESCO, 2023. [Google Scholar] [Crossref]

7. K. Heugh, "Theory and practice in multilingual education," in The SAGE Handbook of Bilingualism, A. Blackledge and A. Creese, Eds. London, UK: SAGE, 2011. [Google Scholar] [Crossref]

8. L. Xue et al., "mT5: A massively multilingual pre-trained text-to-text Transformer," in Proc. 2021 Conf. North Amer. Chapter Assoc. Comput. Linguistics: Hum. Lang. Technol. (NAACL-HLT 2021), 2021, doi: 10.48550/arXiv.2010.11934. [Google Scholar] [Crossref]

9. NLLB Team, "Scaling neural machine translation to 200 languages," Nature, vol. 630, pp. 841–846, 2024, doi: 10.1038/s41586-024-07335-x. [Google Scholar] [Crossref]

10. Y. Liu et al., "Multilingual denoising pre-training for neural machine translation," Trans. Assoc. Comput. Linguistics, vol. 8, pp. 726–742, 2020, doi: 10.1162/tacl_a_00343. [Google Scholar] [Crossref]

11. S. Lankford, "Enhancing neural machine translation of low-resource languages: Corpus development, human evaluation and explainable AI architectures," Ph.D. thesis, arXiv:2403.01580, 2024. [Google Scholar] [Crossref]

12. L. Chang, C. Lin, C. Hsu, and Y. Fan, "Hybrid AI translation framework for minority languages using phrase-based and transformer-based neural machine translation," 2025. [Google Scholar] [Crossref]

13. D. A. Silva et al., "Hybrid rule-based and neural machine translation for Brazilian sign language (Libras)," 2025. [Google Scholar] [Crossref]

14. U. Khaire, R. Chabhaiya, V. Killewale, S. Khandalkar, and P. Ahire, "Hybrid neural machine translation framework for Indian languages using adapter fine-tuning on the PMIndia dataset," 2025. [Google Scholar] [Crossref]

15. D. Bahdanau, K. Cho, and Y. Bengio, "Neural machine translation by jointly learning to align and translate," arXiv:1409.0473, 2015. [Google Scholar] [Crossref]

16. A. Vaswani et al., "Attention is all you need," in Adv. Neural Inf. Process. Syst., vol. 30, 2017, doi: 10.48550/arXiv.1706.03762. [Google Scholar] [Crossref]

17. W. J. Boyd, "Machine translation in the AI era: Comparing previous methods of machine translation with large language models," in Proc. Assoc. Comput. Linguistics (ACL 2025), 2025. [Google Scholar] [Crossref]

18. L. S. Al-Abbas, B. K. Khalaf, O. Alali, and M. Alqaryouti, "Tracking improvement from statistical to neural machine translation: An error-based evaluation of Google Translate," 2025, doi: 10.58256/hnww2b65. [Google Scholar] [Crossref]

19. E. Satir and H. Bulut, "Controlled beam search for neural machine translation using subword units leveraging phrase-based statistical machine translation outputs," Discover Comput., vol. 29, Art. no. 37, 2026, doi: 10.1007/s10791-025-09633-y. [Google Scholar] [Crossref]

20. X. Yu, "Research on English translation system combining statistical machine translation and deep learning," Educ. Res. Rev., vol. 7, no. 4, p. 108, 2025, doi: 10.32629/rerr.v7i4.3864. [Google Scholar] [Crossref]

21. Y. Xiao, L. Wu, J. Guo, J. Li, M. Zhang, T. Qin, and T.-Y. Liu, "A survey on non-autoregressive generation for neural machine translation and beyond," arXiv:2204.09269, 2023. [Google Scholar] [Crossref]

22. C. Raffel et al., "Exploring the limits of transfer learning with a unified text-to-text Transformer," J. Mach. Learn. Res., vol. 21, no. 140, pp. 1–67, 2020. [Google Scholar] [Crossref]

23. Z. Chi et al., "mT6: Multilingual pretrained text-to-text Transformer with translation pairs," in Proc. 2021 Conf. Empirical Methods Nat. Lang. Process. (EMNLP 2021), 2021, doi: 10.48550/arXiv.2104.08692. [Google Scholar] [Crossref]

24. B. Zoph, D. Yuret, J. May, and K. Knight, "Transfer learning for low-resource neural machine translation," in Proc. 2016 Conf. Empirical Methods Nat. Lang. Process. (EMNLP 2016), 2016, doi: 10.48550/arXiv.1604.02201. [Google Scholar] [Crossref]

25. A. Conneau et al., "Unsupervised cross-lingual representation learning at scale," in Proc. 58th Annu. Meeting Assoc. Comput. Linguistics (ACL 2020), 2020, doi: 10.48550/arXiv.1911.02116. [Google Scholar] [Crossref]

26. M. K. Pasupuleti, "Multilingual NLP for low-resource languages using transfer learning," Int. J. Acad. Ind. Res. Innov., vol. 5, no. 5, pp. 452–461, 2025, doi: 10.62311/nesx/rphcr7. [Google Scholar] [Crossref]

27. J. Zhu, H. Sun, and B. Kong, "Improving multilingual English translation performance through T5 and MAML integration," Syst. Soft Comput., 2025, doi: 10.1016/j.sasc.2025.200394. [Google Scholar] [Crossref]

28. T. O. Tafa et al., "Machine translation performance for low-resource languages: A systematic literature review," IEEE Access, 2025, doi: 10.1109/ACCESS.2025.3562918. [Google Scholar] [Crossref]

29. E. Agyei, X. Zhang, A. B. Quaye, V. A. Odeh, and J. R. Arhin, "Dynamic aggregation and augmentation for low-resource machine translation using federated fine-tuning of pretrained Transformer models," Appl. Sci., vol. 15, no. 8, p. 4494, 2025, doi: 10.3390/app15084494. [Google Scholar] [Crossref]

30. R. Kimera, D. Heo, D. N. Rim, and H. Choi, "Data augmentation with back translation for low-resource languages: A case of English and Luganda," in Proc. 2024 8th Int. Conf. Nat. Lang. Process. Inf. Retrieval, 2025, pp. 142–148, doi: 10.1145/3711542.3711594. [Google Scholar] [Crossref]

31. M. A. Faheem, K. T. Wassif, H. Bayomi, and S. M. Abdou, "Improving neural machine translation for low-resource languages through non-parallel corpora: A case study of Egyptian dialect to modern standard Arabic translation," Sci. Rep., vol. 14, Art. no. 2265, 2024, doi: 10.1038/s41598-023-51090-4. [Google Scholar] [Crossref]

32. I. Sel and D. Hanbay, "Efficient adaptation: Enhancing multilingual models for low-resource language translation," Mathematics, vol. 12, no. 19, p. 3149, 2024, doi: 10.3390/math12193149. [Google Scholar] [Crossref]

33. M. Javed et al., "A transformer-based re-ranking model for low-resource neural machine translation with bilingual curriculum learning and contrastive re-ranking," 2025. [Google Scholar] [Crossref]

34. Y. Wu, J. Zhang, and colleagues, "Hybrid neural machine translation and large language model scheduling for improved translation quality," 2025. [Google Scholar] [Crossref]

35. W. Nekoto et al., "Participatory research for low-resourced machine translation: A case study in African languages," arXiv:2010.02353, 2020, doi: 10.48550/arXiv.2010.02353. [Google Scholar] [Crossref]

36. Y. Moslem, A. K. Wassie, and A. G. Abebe, "AfriNLLB: Efficient translation models for African languages," in Proc. 7th Workshop African Nat. Lang. Process. (AfricaNLP 2026), 2026, doi: 10.48550/arXiv.2602.09373. [Google Scholar] [Crossref]

37. J. O. Alabi et al., "AFRIDOC-MT: Document-level MT corpus for African languages," in Proc. 2025 Conf. Empirical Methods Nat. Lang. Process. (EMNLP 2025), 2025, doi: 10.48550/arXiv.2501.06374. [Google Scholar] [Crossref]

38. C. C. Emezue and B. F. P. Dossou, "MMTAfrica: Multilingual machine translation for African languages," in Proc. 6th Conf. Mach. Transl. (WMT 2021), 2021, pp. 398–411, doi: 10.48550/arXiv.2204.04306. [Google Scholar] [Crossref]

39. S. Issaka et al., "The African Languages Lab: A collaborative approach to advancing low-resource African NLP," arXiv:2510.05644, 2025, doi: 10.48550/arXiv.2510.05644. [Google Scholar] [Crossref]

40. Y. J. Waliya and M. M. Okon, "Metadata analysis of the generative AI usefulness for African languages," Preprint, 2025. [Google Scholar] [Crossref]

41. A. A. Tapo et al., "Neural machine translation for extremely low-resource African languages: A case study on Bambara," arXiv:2011.05284, 2020, doi: 10.48550/arXiv.2011.05284. [Google Scholar] [Crossref]

42. C. Gitau and V. Marivate, "Textual augmentation techniques applied to low resource machine translation: Case of Swahili," arXiv:2306.07414, 2023, doi: 10.48550/arXiv.2306.07414. [Google Scholar] [Crossref]

• Block-Based Programming for Education: A Comprehensive Analysis of Visual Programming Environments in K-12 Learning
• Management of Academic Libraries and Client Satisfaction Towards Digital Utilization: Basis for Monitoring Library Operations in SOCCSKSARGEN Region.
• Revenue Leakages in TPA Insurance Claims and Corporate Claims: An Institutional Overview of Aster Prime Hospital, Hyderabad
• Technology and Innovation in Hospitality and Tourism: A Management Perspective
• Geospatial Distribution of Tarok Sacred Grove of Langtang North and Langtang South Local Government Areas