AIAny
Icon for item

Uzbek Language Corpus

Provides 5.6 million cleaned Uzbek documents—about 6.06 billion words across Latin and Cyrillic—for language-model pretraining and NLP. It combines books, archives, legal texts, web pages and public Telegram posts with source, topic, domain and register annotations.

Introduction

For Uzbek, a large corpus has to bridge two writing systems as well as very different kinds of text. This collection includes both Latin and Cyrillic material, making its value more than sheer scale: it gives language-model training data a broader view of how Uzbek is written across scripts and settings.

What the Coverage Makes Possible

The corpus brings together books, periodicals, archival records, legal acts, Wikipedia, web pages and public Telegram posts. That mix can support general Uzbek language-model pretraining and NLP research, while document-level source, topic, domain, register and script fields help researchers inspect or filter the data. Personal identifiers are replaced with typed placeholders, and exact duplicate texts are removed.

Great Fit If…

It suits teams training Uzbek models or studying Uzbek text across scripts and domains. Look elsewhere if you need a balanced conversational dataset or verified factual labels: books and archives dominate the words, conversational material is comparatively limited, and roughly 1.2 million documents came from OCR and may retain recognition errors. Language-model-generated annotations can also be imperfect.

Information

  • Websitehuggingface.co
  • OrganizationsMinistry of Digital Technologies of the Republic of Uzbekistan
  • Published date2026/11/18

Categories

More Items

Scales verifiable gameplay data by having a Builder and visual and coding players co-develop web games. Progress-weighted training improves agent performance on both its own benchmark and an external game benchmark.

Hugging Face

Provides executable software-engineering repair tasks for evaluating coding agents. Its 2,955 synthetic instances span 53 Python repositories and include bug-injection patches, reference repairs, and test information.

Hugging Face

Pairs multilingual documents with JSON Schemas and target extractions for training and evaluating structured information extraction. Its held-out document types make it possible to test generalization beyond the tasks seen during training.