This dataset consists of clean, structured, and filtered Somali language text compiled from various online sources. It is designed for training and fine-tuning Somali language models (LLMs) and supporting natural language processing (NLP) research for the Somali language.

Dataset Details

Language: Somali (so) Format: JSON lines (.jsonl) Data Structure: Each record has a single text field containing a cleaned paragraph.

Sources… See the full description on the dataset page: https://huggingface.co/datasets/maanka2/somali-web-corpus.