Uzbek NER Gold: A Gold-Standard Dataset for Uzbek Named Entity Recognition

Uzbek NER Gold is a gold-standard token-level named entity recognition dataset for the Uzbek language. The dataset contains 59,569 annotated token rows from 4,176 sentences and is provided in UTF-8 TSV format. Each row includes a sentence identifier, token order, token text, BIO-formatted NER tag, and POS/category label. The annotation schema covers eight named entity types: Person (PER), Organization (ORG), Location (LOC), Miscellaneous (MISC), Money (MONEY), Number (NUMERIC), Date/Time (TEMPORAL), and Work (WORK). The released dataset includes 9,250 named entity spans and 17 BIO labels plus the outside label O. The text was normalized to Uzbek Latin script, and Uzbek apostrophe variants were standardized. This dataset is intended for Uzbek NLP research and development, including named entity recognition, token classification, information extraction, and evaluation of Uzbek language models.

Hali tarjima qilinmagan