Parameter-Efficient Few-Shot Sentiment Analysis Using LoRA-Enhanced Transformers

Nurudeen Jibrin

doi:10.14710/jmasif.17.1.81053

DOI: https://doi.org/10.14710/jmasif.17.1.81053

Parameter-Efficient Few-Shot Sentiment Analysis Using LoRA-Enhanced Transformers

Nurudeen Jibrin

Department of Computer Science, Baze University, Abuja, Nigeria, Nigeria

Received: 28 Dec 2025; Revised: 1 Mar 2026; Accepted: 6 Apr 2026; Published: 13 Apr 2026.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

BibTex Citation Data :

@article{JMASIF81053,
    author = {Nurudeen Jibrin},
    title = {Parameter-Efficient Few-Shot Sentiment Analysis Using LoRA-Enhanced Transformers},
    journal = {Jurnal Masyarakat Informatika},
  volume = {17},
    number = {1},
    year = {2026},
    keywords = {Low-Rank Adaptation; Sentiment Analysis; Hausa Language; Natural Language Processing; Few-Shot Learning},
    abstract = {Sentiment analysis in low-resource languages remains constrained by limited annotated data and the computational expense of fine-tuning large language models. This study introduces a parameter-efficient framework that applies Low-Rank Adaptation (LoRA) to lightweight transformers (AfriBERTa, DistilBERT, and MiniLMv2) for Hausa sentiment analysis on the NaijaSenti dataset. The approach addresses three key challenges: effective few-shot learning, robustness under severe data scarcity (100-1,000 training samples), and mitigation of language-specific linguistic errors. Results show that AfriBERTa-LoRA achieved 69.0% accuracy with only 4.8 percentage points below a fully fine-tuned XLM-RoBERTa baseline. Concurrently, the model utilised only 1.06% of trainable parameters (94-fold reduction) and decreased GPU memory requirements by approximately 50%. Performance scales efficiently with data volume, narrowing the gap to full fine-tuning as training samples increase. Linguistic error analysis identified four Hausa-specific failure modes accounting for 71.5% of misclassifications, with targeted mitigations yielding an 8.7 percentage point absolute reduction in error rate. This work establishes LoRA as a highly effective paradigm for low-resource Natural Language Processing (NLP), delivering near-state-of-the-art Hausa sentiment analysis at minimal cost and offering a replicable pipeline for other underrepresented African languages.},
   issn = {2777-0648},   pages = {146--159}  doi = {10.14710/jmasif.17.1.81053},
    url = {https://ejournal.undip.ac.id/index.php/jmasif/article/view/81053}
}

Citation Format:

Abstract

Sentiment analysis in low-resource languages remains constrained by limited annotated data and the computational expense of fine-tuning large language models. This study introduces a parameter-efficient framework that applies Low-Rank Adaptation (LoRA) to lightweight transformers (AfriBERTa, DistilBERT, and MiniLMv2) for Hausa sentiment analysis on the NaijaSenti dataset. The approach addresses three key challenges: effective few-shot learning, robustness under severe data scarcity (100-1,000 training samples), and mitigation of language-specific linguistic errors. Results show that AfriBERTa-LoRA achieved 69.0% accuracy with only 4.8 percentage points below a fully fine-tuned XLM-RoBERTa baseline. Concurrently, the model utilised only 1.06% of trainable parameters (94-fold reduction) and decreased GPU memory requirements by approximately 50%. Performance scales efficiently with data volume, narrowing the gap to full fine-tuning as training samples increase. Linguistic error analysis identified four Hausa-specific failure modes accounting for 71.5% of misclassifications, with targeted mitigations yielding an 8.7 percentage point absolute reduction in error rate. This work establishes LoRA as a highly effective paradigm for low-resource Natural Language Processing (NLP), delivering near-state-of-the-art Hausa sentiment analysis at minimal cost and offering a replicable pipeline for other underrepresented African languages.

Fulltext View|Download Email colleagues

Keywords: Low-Rank Adaptation; Sentiment Analysis; Hausa Language; Natural Language Processing; Few-Shot Learning

Article Metrics:

Article Info

Section: Research Article

Language : EN

In Vol 17, No 1 (2026): May 2026 (Ongoing)

Most cited articles

Enterprise Architecture Model untuk Aplikasi Government ANALISA PERFORMA METODE COSINE DAN JACARD PADA PENGUJIAN KESAMAAN DOKUMEN PENYELESAIAN MASALAH JOB SHOP MENGGUNAKAN ALGORITMA GENETIKA Studi Komparasi Ekstraksi Fitur pada Pengenalan Wajah Menggunakan Principal Component Analysis (PCA) dan Wavelet Daubechies ALGORITMA BACK PROPAGATION NEURAL NETWORK UNTUK PENGENALAN POLA KARAKTER HURUF JAWA More cited articles

B. I. Adekunle, E. C. Chukwuma-Eke, E. D. Balogun, and K. O. Ogunsola, ‘Sentiment Analysis for Customer Behavior Insights: A Natural Language Processing Approach to Business Decision-Making’, IJSSER, vol. 3, no. 1, pp. 272–282, 2024, doi: 10.54660/IJSSER.2024.3.1.272-282
T. Joseph, ‘Natural Language Processing (NLP) for Sentiment Analysis in Social Media’, IJCE, vol. 6, no. 2, pp. 35–48, July 2024, doi: 10.47941/ijce.2135
K. Ronny Mabokela, M. Primus, and T. Celik, ‘Advancing sentiment analysis for low-resourced african languages using pre-trained language models’, PLoS One, vol. 20, no. 6, p. e0325102, June 2025, doi: 10.1371/journal.pone.0325102
J. McGiff and N. S. Nikolov, ‘Overcoming Data Scarcity in Generative Language Modelling for Low-Resource Languages: A Systematic Review’, 2025, arXiv. doi: 10.48550/ARXIV.2505.04531
S. A. Abdulmumin, ‘Hausa language communication and development’, presented at the 1st International Conference on Communication, Media, Insecurity, and Development: Issues, Challenges and Way Forward, Dept. Mass Commun., Ahmadu Bello Univ., Zaria, Nigeria, Available: https://www.researchgate.net/publication/386902803_HAUSA_LANGUAGE_COMMUN CATION_AND_DEVELOPMENT
A.-U. Sani and M. U. Mustapha, ‘Global Growing Impact of Hausa and the Need for its Documentation’, Contemporary Journal of Language and Literature, vol. 1(1), pp. 16–34, 2018, Available: https://www.researchgate.net/publication/344409625_Global_Growing_Impact_of_Hausa_and_the_Need_for_its_Documentation
S. H. Muhammad et al., ‘NaijaSenti: A Nigerian Twitter Sentiment Corpus for Multilingual Sentiment Analysis’, June 18, 2022, arXiv: arXiv:2201.08277. doi: 10.48550/arXiv.2201.08277
A. Yusuf, A. Sarlan, K. U. Danyaro, and A. S. B. A. Rahman, ‘Fine-tuning Multilingual Transformers for Hausa-English Sentiment Analysis’, in 2023 13th International Conference on Information Technology in Asia (CITA), Kota Samarahan, Malaysia: IEEE, Aug. 2023, pp. 13–18. doi: 10.1109/CITA58204.2023.10262742
I. Inuwa-Dutse, ‘NaijaNLP: A Survey of Nigerian Low-Resource Languages’, Mar. 06, 2025, arXiv: arXiv:2502.19784. doi: 10.48550/arXiv.2502.19784
F. M. Adam, A. Y. Zandam, and I. Inuwa-Dutse, ‘Detection and Analysis of Offensive Online Content in Hausa Language’, Apr. 26, 2024, In Review. doi: 10.21203/rs.3.rs-4266465/v2
N. A. Sharma, A. B. M. S. Ali, and M. A. Kabir, ‘A review of sentiment analysis: tasks, applications, and deep learning techniques’, Int J Data Sci Anal, vol. 19, no. 3, pp. 351–388, Apr. 2025, doi: 10.1007/s41060-024-00594-x
L. Wang et al., ‘Parameter-efficient fine-tuning in large language models: a survey of methodologies’, Artif Intell Rev, vol. 58, no. 8, p. 227, May 2025, doi: 10.1007/s10462-025-11236-4
Y. Mao et al., ‘A survey on LoRA of large language models’, Front. Comput. Sci., vol. 19, no. 7, p. 197605, July 2025, doi: 10.1007/s11704-024-40663-9
X. Li and A. Kim, ‘A Study to Evaluate the Impact of LoRA Fine-tuning on the Performance of Non-functional Requirements Classification’, 2025, arXiv. doi: 10.48550/ARXIV.2503.07927
Z. Liu, J. Lyn, W. Zhu, X. Tian, and Y. Graham, ‘ALoRA: Allocating Low-Rank Adaptation for Fine-tuning Large Language Models’, Apr. 15, 2024, arXiv: arXiv:2403.16187. doi: 10.48550/arXiv.2403.16187
Y. Wang, Q. Yao, J. T. Kwok, and L. M. Ni, ‘Generalizing from a Few Examples: A Survey on Few-shot Learning’, ACM Comput. Surv., vol. 53, no. 3, pp. 1–34, May 2021, doi: 10.1145/3386252
S. H. Muhammad et al., ‘SemEval-2023 Task 12: Sentiment Analysis for African Languages (AfriSenti-SemEval)’, May 01, 2023, arXiv: arXiv:2304.06845. doi: 10.48550/arXiv.2304.06845
T. Dettmers, A. Pagnoni, A. Holtzman, and L. Zettlemoyer, ‘QLoRA: Efficient Finetuning of Quantized LLMs’, 2023, arXiv. doi: 10.48550/ARXIV.2305.14314
H. Zhang, X. Mu, H. Yan, L. Ren, and J. Ma, ‘A survey of online video advertising’, WIREs Data Min & Knowl, vol. 13, no. 2, p. e1489, Mar. 2023, doi: 10.1002/widm.1489
V. Lialin, V. Deshpande, X. Yao, and A. Rumshisky, ‘Scaling Down to Scale Up: A Guide to Parameter-Efficient Fine-Tuning’, 2023, arXiv. doi: 10.48550/ARXIV.2303.15647

Last update:

No citation recorded.

Last update: 2026-04-27 13:19:56

No citation recorded.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

The authors who submit the manuscript must understand that the article's copyright belongs to the author(s) if accepted for publication. However, the author(s) grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution-ShareAlike 4.0 International License. Authors should also understand that their article (and any additional files, including data sets, and analysis/computation data) will become publicly available once published under that license. By submitting the manuscript to Jmasif, the author(s) agree with this policy. No special document approval is required.

The author(s) guarantee that:

their article is original, written by the mentioned author(s),
has never been published before,
does not contain statements that violate the law, and
does not violate the rights of others, is subject to copyright held exclusively by the author(s), is free from the rights of third parties, and the necessary written permission to quote from other sources has been obtained by the author(s).

The author(s) retain all rights to the published work, such as (but not limited to) the following rights:

Copyright and other proprietary rights related to the article, such as patents,
The right to use the substance of the article in its own future works, including lectures and books,
The right to reproduce the article for its own purposes,
The right to archive all versions of the article in any repository, and
The right to enter into separate additional contractual arrangements for the non-exclusive distribution of published versions of the article (for example, posting them to institutional repositories or publishing them in a book), acknowledging its initial publication in this journal (Jurnal Masyarakat Informatika).

Suppose the article was prepared jointly by more than one author. Each author submitting the manuscript warrants that all co-authors have given their permission to agree to copyright and license notices (agreements) on their behalf and notify co-authors of the terms of this policy. Jmasif will not be held responsible for anything arising because of the writer's internal dispute. Jmasif will only communicate with correspondence authors.

Authors should also understand that their articles (and any additional files, including data sets and analysis/computation data) will become publicly available once published. The license of published articles (and additional data) will be governed by a Creative Commons Attribution-ShareAlike 4.0 International License. Jmasif allows users to copy, distribute, display and perform work under license. Users need to attribute the author(s) and Jmasif to distribute works in journals and other publication media. Unless otherwise stated, the author(s) is a public entity as soon as the article is published.