- Abdulhamid Abubakar1, Gilbert I. O. Aimufua2, Muhammed A. Abubakar3
- DOI: 10.5281/zenodo.21848915
- SSR Journal of Artificial Intelligence (SSRJAI)
The proliferation of propaganda through
digital platforms threatens information integrity globally. Computational tools
for fine-grained persuasion technique detection remain concentrated on English
and a few high-resource European languages. Hausa, a Chadic language spoken by
over 120 million people across West and Central Africa, lacks systems capable of
identifying specific persuasion strategies like loaded language or appeal to
authority. This systematic review, conducted following PRISMA 2020 guidelines,
synthesises research at the intersection of fine-grained propaganda detection,
parameter-efficient transfer learning, cross-lingual knowledge transfer, and
African natural language processing (NLP). Thirty-seven studies (2019-2026)
were retrieved from IEEE Xplore, ScienceDirect, Scopus, Web of Science, and the
ACL Anthology based on strict inclusion criteria. The evidence reveals three
primary patterns: (a) fine-grained propaganda detection features mature
taxonomies and transformer-based methods, but its empirical base excludes
African languages; (b) parameter-efficient methods like Low-Rank Adaptation
LoRA match full fine-tuning performance but remain unevaluated for span-level
technique detection; and (c) Hausa NLP has largely advanced through binary and
coarse classification, leaving a critical gap in fine-grained analysis. A
coverage analysis confirms that no existing study addresses fine-grained
propaganda detection for Hausa. This review consolidates the evidence in nine
tables and seven figures, maps the resource landscape, profiles adaptation
trade-offs, and proposes a five-year research roadmap. The roadmap prioritises
adapter benchmarking and corpus development (2026-2028), cross-lingual transfer
with cultural adaptation (2027-2029), and deployable, resource-efficient
systems for African fact-checking organizations (2029-2031). The findings demonstrate
that while the technical building blocks for fine-grained propaganda detection
in Hausa exist, the intersection remains completely unexplored, presenting a
significant opportunity for future research.
Keywords: Fine-grained propaganda detection; cross-lingual transfer
learning; parameter-efficient fine-tuning; Hausa; low-resource languages;
African language NLP; systematic literature review.
