article · International Journal of Mathematics Statistics and Computer Science
Training machine reading comprehension models on Swahili poses challenges due to datasets combining both extractive and abstractive responses. Using the multilingual T5 architecture on the Kencorpus Swahili Question Answering Dataset, a hierarchical curriculum learning method separated the data into extractive and abstractive subsets. Training progressed sequentially from English structural transfer to Swahili extractive alignment, ending with abstractive fine-tuning, alongside structure-aware input scaffolding for long texts. The base model achieved a SacreBLEU score of 48.99 on extractive tasks. For abstractive question answering, the model scored 15.52 on BLEU but reached a BERTScore F1 of 77.21 per cent, showing strong semantic understanding despite lower exact-word overlap. The results demonstrate that while multilingual models capture Swahili question-answering logic well, closing the fluency gap in abstractive answering will require larger, targeted datasets.
Natural language processing tools for African languages often struggle with complex reasoning tasks. Demonstrating that multilingual models can achieve high semantic accuracy in Swahili question answering shows significant progress for automated comprehension in the language. However, the identified fluency limitations in abstractive answers highlight where further data collection and specialised training are necessary before these systems can produce human-level generated text.
This work represents early-stage research relevant to software developers and language technology firms creating Swahili search engines, customer support bots, and knowledge retrieval systems. While the extractive model demonstrates strong readiness for information retrieval tasks, abstractive generation remains experimental due to fluency deficits. Commercial deployment in open-ended generative applications will require the creation of larger, targeted datasets to improve lexical output.
AI-generated from the published abstract. Always read the original work before citing.
The introduction of the Kencorpus Swahili Question Answering Dataset (KenSwQuAD) presents a compelling opportunity to advance Machine Reading Comprehension (MRC) for the Swahili language. However, the mixed composition of this dataset—comprising 67.6\% extractive and 32.4\% abstractive answers—introduces substantial hurdles for standard training pipelines. In this study, we benchmark the performance of the multilingual T5 (mT5) architecture on KenSwQuAD, deploying a robust Hierarchical Curriculum Learning strategy. We introduce a novel data restructuring technique that algorithmically partitions the corpus into distinct extractive and abstractive subsets, thereby enabling meticulously phased fine-tuning. By sequentially progressing from structural transfer via the English SQuAD, to morphological alignment on the Swahili extractive subset, and culminating in abstractive refinement, we establish that the mT5-base model can achieve a SacreBLEU score of 48.99 on extractive tasks. For abstractive reasoning, whilst the BLEU score is comparatively modest (15.52), the model attains a remarkable BERTScore F1 of 77.21\%, reflecting a profound degree of semantic comprehension. Furthermore, we evaluate the utility of structure-aware input formatting (context scaffolding) for successfully navigating extensive narrative contexts. Ultimately, our findings indicate that whilst modern transformer architectures can successfully internalise Swahili QA logic, a persistent fluency gap characterises the abstractive domain. This gap underscores the urgent necessity for more expansive, targeted datasets to reconcile semantic accuracy with lexical precision.
This page summarises published work. The authoritative version sits with the publisher.
DOI: 10.59543/14d9kq02
Is something wrong with this record? Report it or request removal.
Discussion
Have you built on this work, tried to replicate it, or seen it applied in practice? Share what you know. Verified researchers and MARATTO™ domain experts can open a discussion, and any member can reply. Contributions are reviewed before they appear.
No discussion yet. Open the first thread.
New to MARATTO™? Create a free account.
