metadata

dataset_info:
  features:
    - name: text
      dtype: string
  splits:
    - name: train
      num_bytes: 827986534
      num_examples: 47982455
  download_size: 429416941
  dataset_size: 827986534
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: apache-2.0
task_categories:
  - token-classification
language:
  - uk
size_categories:
  - 10M<n<100M
tags:
  - silver-standard
  - ukrainian
  - NER

UberText-NER-Silver

UberText-NER-Silver is a silver-standard named entity recognition (NER) dataset for the Ukrainian language. It was automatically annotated using a high-performance model trained on NER-UK 2.0 and covers over 2.5 million social media and web sentences. The dataset significantly expands the coverage of underrepresented entity types and informal domains.

Dataset Summary

Total Sentences: 2,573,205
Total Words: 45,489,533
Total Entity Spans: 4,393,316
Entity Types (13): PERS, ORG, LOC, DATE, TIME, JOB, MON, PCT, PERIOD, DOC, QUANT, ART, MISC
Format: IOB-style, token-level annotations

Source

Texts were taken from the UberText 2.0 corpus social media part, filtered and preprocessed for noise reduction and duplication. The dataset includes both entity-rich and entity-free content to improve model generalization.

Example Usage

from datasets import load_dataset

dataset = load_dataset("lang-uk/UberText-NER-Silver", split="train")
print(dataset[0])

Applications

Training large-scale NER models for Ukrainian
Improving performance in low-resource and informal text domains
Cross-lingual or transfer learning experiments

Authors

Vladyslav Radchenko, Nazarii Drushchak