A Text Summarization Model for the Yorùbá Language

Text summarization is important in this era of electronic data explosion. While there exist different models for text summarization in High-resource languages like English, there is no automatic text summarization model for the Yorùbá language. Leveraging the large language and pre-trained language models, a text summarization technique for the Yorùbá language which fine-tunes the mT5 model with a monolingual Yorùbá dataset is proposed. The model implemented is a transformer-based sequence-to-sequence abstractive text summarization model which trained the mT5 model using the XL-Sum Yorùbá bilingual dataset. We assess the effectiveness of the model in summarising Yorùbá text, providing insight into its ability to extract valuable information from a large amount of textual input. The Yorùbá text summarization model we developed achieved the following ROUGE scores, 41.27 for ROUGE-1, 19.77 for ROUGE-2, 30.57 for ROUGE-L, and 30.59 for RogueLsum. The outcome demonstrates reasonable Yorùbá text summarization, setting a base standard for further text summarization models for the Yorùbá language.