EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

Interpreting isiZulu Idioms with Large Language Models: A Comparative Study of Zero-Shot Prompting, Retrieval-Augmented Generation, and Supervised Fine-Tuning

Nontokozo Buhle Ntandoyenkosi Manukuza, Vukosi Marivate, Joseph Sefara, Kwanele R, Sthembiso Mkhwanazi

Journal of the Digital Humanities Association of Southern Africa (DHASA) · 20.11.2026 · 7 (2) · 151-166

DOI: 10.55492/dhasa.v7i02.7708 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of the Digital Humanities Association of Southern Africa (DHASA), online erschienen am 20.11.2026. Verfasst von Nontokozo Buhle Ntandoyenkosi Manukuza und 4 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Idiom interpretation requires cultural knowledge, context, and pragmatic understanding that cannot be inferred from literal word meanings alone. This challenge is particularly important for isiZulu, a richly inflected Nguni Bantu language with limited representation in the web-scale corpora used to train modern large language models (LLMs). We compare three approaches to isiZulu idiom interpretation (zero-shot prompting, retrieval-augmented generation (RAG) with glossary evidence, and supervised fine-tuning on a purpose-built, context-rich corpus) across five open-weight LLMs, evaluating them on 323 held-out idioms. Performance is evaluated using automated metrics and human assessment by expert isiZulu speakers. Supervised fine-tuning provides the strongest overall improvement, increasing the mean human evaluation score from 2.61 to 3.68 out of 5. RAG also improves interpretation quality, although its gains are smaller and are not fully captured by conventional surface-level metrics. Across the fifteen model--condition systems, COMET aligns best with human assessment, whereas embedding cosine similarity is the only metric that does not align positively with it, suggesting that this metric should be validated against human judgement before being used to evaluate idiomatic interpretation in low-resource languages. We contribute the isiZulu Idiom Dataset, a morphology-aware retrieval pipeline, and a reusable protocol for evaluating LLM-based idiom interpretation in low-resource African languages.

Zitieren

APA

Manukuza, N. B. N., Marivate, V., Sefara, J., R, K., & Mkhwanazi, S. (2026). Interpreting isiZulu Idioms with Large Language Models: A Comparative Study of Zero-Shot Prompting, Retrieval-Augmented Generation, and Supervised Fine-Tuning. Journal of the Digital Humanities Association of Southern Africa (DHASA), 7(2), 151-166. https://doi.org/10.55492/dhasa.v7i02.7708

Vancouver

Manukuza NBN, Marivate V, Sefara J, R K, Mkhwanazi S. Interpreting isiZulu Idioms with Large Language Models: A Comparative Study of Zero-Shot Prompting, Retrieval-Augmented Generation, and Supervised Fine-Tuning. Journal of the Digital Humanities Association of Southern Africa (DHASA). 2026;7(2):151-166. doi:10.55492/dhasa.v7i02.7708

BibTeX
@article{Manukuza2026,
  author = {Manukuza, Nontokozo Buhle Ntandoyenkosi and Marivate, Vukosi and Sefara, Joseph and R, Kwanele and Mkhwanazi, Sthembiso},
  title = {Interpreting isiZulu Idioms with Large Language Models: A Comparative Study of Zero-Shot Prompting, Retrieval-Augmented Generation, and Supervised Fine-Tuning},
  journal = {Journal of the Digital Humanities Association of Southern Africa (DHASA)},
  year = {2026},
  volume = {7},
  number = {2},
  pages = {151--166},
  doi = {10.55492/dhasa.v7i02.7708}
}
Kostenlos starten →