Skip to content

Dataset · ontonotes

Multilingual OntoNotes 5.0 corpus for CoNLL-2012

ontonotes/conll2012_ontonotesv5

Large English, Chinese, and Arabic corpus with manual syntactic and semantic annotation.

What's inside

The final version of the OntoNotes v5.0 corpus, extended for the CoNLL-2012 competition and taken from the Mendeley Data repository. Contains multi-genre texts in English, Chinese, and Arabic manually annotated with syntactic, semantic, and discourse information (from 10 to 100 thousand examples). Includes training, validation, and test splits of versions v4, v9, as well as a corrected v12 version for English only.

What you can do with it

  • Train token labeling models (NER, part-of-speech, syntactic parsing)
  • Evaluate NLP model performance on the standard CoNLL-2012 benchmark
  • Train multilingual text analysis systems for English, Chinese, and Arabic
Tasks
Token classification
Languages
Arabic, English, Chinese
Size
10K–100K rows
Downloads this month
2K
Likes
47

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("ontonotes/conll2012_ontonotesv5")