Tashkent, Uzbekistan (UzDaily.uz) — Uzbekistan’s Ministry of Digital Technologies has published an open Uzbek-language text corpus containing 5.6 million documents and around 6 billion words on the Hugging Face platform, the ministry’s press service said.
The corpus includes Uzbek-language materials in both Latin and Cyrillic scripts. Its contents include digitised books, newspaper articles, archival materials, legal documents, Wikipedia articles, web pages and publicly available Telegram posts.
The collected texts underwent specialised processing and quality screening. The materials were also classified by industry and subject area.
Personal data was replaced with special markers, including [PHONE] and [NAME].
The corpus was developed as part of efforts to implement an instruction issued by President Shavkat Mirziyoyev on 21 October 2025. The instruction was given at a videoconference meeting on accelerating the use of artificial intelligence in public administration, economic sectors and regions.
At the meeting, Mirziyoyev stressed the need for AI systems to understand the Uzbek language fully.
The project aims to expand the digital language database and improve AI models’ ability to process Uzbek text, understand the language and generate high-quality content in Uzbek.
The ministry is now working to make the corpus available to developers of leading large language models (LLMs). It said cooperation with the Qwen model had begun.
The open corpus is available on Hugging Face: https://huggingface.co/datasets/Digitaluz/uzbek-language-corpus