Synthesising a corpus of Gaelic traditional narrative with cross-lingual text expansion

Advances in large language modelling have disproportionately benefited high-resource languages due to their vastly greater training data reserves. This paper proposes a novel cross-lingual text expansion (XLTE) technique using multilingual large language models (MLLMs) to mitigate data sparsity in low-resource languages (LRLs). We apply XLTE to the domain of traditional Scottish Gaelic storytelling, aiming to generate coherent, long-form narrative texts by leveraging MLLMs’ cross-lingual capabilities. The effectiveness of this technique is demonstrated using GPT-4o, with supervised fine-tuning (SFT) providing a $57.2\%$ reduction in perplexity and decreased neologism over baseline models. Despite these improvements, qualitative analyses reveal stylistic divergences between synthesised and genuine data. Nevertheless, XLTE offers a promising, scalable method for synthesising training sets in other languages and domains, opening avenues for further improvements in low-resource language modelling.