AKU: An African Journal of Contemporary Research

This study evaluates the performance of artificial intelligence (AI) translation tools in rendering Igbo into English, with the goal of identifying their strengths and weaknesses and determining their potential for wider application to other low-resource languages. The motivation lies in the urgent need to ensure that African languages, often underrepresented in digital corpora, are effectively supported in emerging language technologies. To achieve this, forty-four Igbo inputs were constructed, covering lexical items, simple sentences, complex sentences, interrogatives, imperatives, and proverbs. These were translated using two widely accessible AI systems, and their outputs were systematically compared with linguist-validated reference translations for accuracy, fluency, and cultural appropriateness. The analysis shows that while the tools perform well with isolated words and straightforward sentences, they encounter difficulties with context-dependent expressions, figurative language, and proverbs, often leading to semantic distortion or cultural loss. Interestingly, one incidental observation was that the tools occasionally recognized tonal distinctions, although tone was not the focus of this research. The study contributes to the field of language technology by providing a diagnostic evaluation of AI translation in Igbo, establishing a replicable methodology for assessing other low-resource languages, and highlighting the need for richer Igbo corpora, tone-sensitive datasets, and culturally grounded training materials. Ultimately, it offers practical insights into how human linguistic expertise and AI systems can complement one another in building more reliable multilingual translation tools.

Open Full Text (PDF)