Currency
  • Loading...
Weather
  • Loading...
Air Quality (AQI)
  • Loading...

The Ministry of Digital Technologies of Uzbekistan announced that it has developed a large open corpus of Uzbek-language text data and published it on the Hugging Face platform for open use. The corpus contains 5.6 million documents and nearly 6 billion words; its materials are compiled in both the Latin and Cyrillic scripts of Uzbek and were created under the implementation of a directive issued at a video conference on October 21, 2025.

Key Takeaways: The corpus is available for open use on the Hugging Face platform; it contains 5.6 million documents and about 6 billion words; the texts are compiled in Latin and Cyrillic scripts; and the collection includes scanned books, newspapers and archival materials, legal documents, Wikipedia articles, web pages, and open Telegram posts.

The initiative was carried out within the framework of a directive issued at a video conference on October 21, 2025, chaired by President Shavkat Mirziyoyev, on the accelerated introduction of artificial intelligence technologies in public administration, economic sectors, and regions. At that time, the Uzbek leadership emphasized that AI programs must, first of all, understand the Uzbek language at an advanced level in order to work effectively, and set the task of creating an “Uzbek Language Corpus.” This step can be seen as a foundation for expanding language resources and strengthening the local AI ecosystem.

The collected texts underwent special processing, were filtered by quality, and classified by field and topic. In addition, personal data was replaced with special markers such as [PHONE] and [NAME], which helps make open data distribution safer. The corpus includes scanned books, newspapers and archival materials, legal documents, Wikipedia articles, web pages, and open Telegram posts.

Publication on an open platform gives researchers, developers, and companies working in artificial intelligence access to a large volume of Uzbek-language texts. According to the ministry, work is currently underway to provide the corpus to developers of leading large language models (LLMs); in particular, collaborative work with the Qwen model has begun. This indicates that the corpus is intended not only for data storage but also for practical model development.

The main goal of the initiative is to expand the digital database in the Uzbek language and to improve the ability of AI models to understand, process, and generate high-quality texts in Uzbek. In the future, models that better support the Uzbek language are expected to emerge, and AI services are expected to become more convenient for local users.

Source: www.gazeta.uz