This dataset is a highly cleaned, deduplicated, and normalized corpus of Turkish text. It features a diverse collection of informative, encyclopedic, and journalistic content.
The dataset is primarily designed for Continued Pre-Training (CPT) of Large Language Models (LLMs) to enhance their Turkish language capabilities and domain knowledge. It is also an excellent foundational corpus for generating synthetic… See the full description on the dataset page: https://huggingface.co/datasets/mehmettozlu/clean-tr-dataset.
