Congratulations to Frazer Nyambe, an MSc in Computer Science student in the DataLab Research Group, who won two prizes at the School of Natural and Applied Sciences Postgraduate Seminar Week, held at the University of Zambia on September 14 and 15, 2026:
- 3rd prize, MSc Oral Presentations (Stream B)
- Joint 2nd prize, MSc Poster Presentations (Stream B)
Frazer's poster, Bridging Zambia's Wikipedia Knowledge Gap using Hybrid Retrieval Augmented Generation and Large Language Models, co-authored with Lighton Phiri, starts from the low level of local participation in Wikipedia in Zambia and the gaps in Zambian knowledge that result. It compares three AI workflows for producing accurate, grounded Zambia-specific content: base large language models, a Llama 3.1 model fine-tuned with QLoRA on information from Zambian universities and local councils, and a hybrid retrieval-augmented generation (RAG) system. Preliminary results from his Open WebUI RAG prototype show that optimised chunking and vector search reduce hallucinations and improve the accuracy of responses.
The study has three objectives: to design and configure the three human–AI content generation workflows in a local Open WebUI environment, using a curated Zambia-specific dataset; to develop an evaluation framework that benchmarks them on factual accuracy, text grounding, faithfulness and contextual recall; and to evaluate, through a controlled within-subjects experiment with student Wikipedia editors, how the best-performing platform affects editing efficiency, content depth and user experience.
The work is part of the Wikipedia in the Global South project and builds on Frazer's ICICT 2025 paper on automatic creation of Wikipedia articles about Zambia. He rehearsed his presentation with the research group on September 11, 2026.