Skip to content

Dataset · eriktks

CoNLL-2003 Classic NER Corpus

eriktks/conll2003

English dataset for named entity recognition: people, locations, organizations, and more.

What's inside

Contains data from the CoNLL-2003 shared task on language-independent NER. The files have four space-separated columns: word, POS tag, syntactic chunk tag, and NER tag. The NER markup uses IOB2 and four entity types: persons, locations, organizations, and miscellaneous. English language, size between 10K and 100K.

What you can do with it

  • Training a token-classification model for NER
  • Testing an NER pipeline on a classic benchmark
  • Comparing IOB1 and IOB2 labeling schemes
Tasks
Token classification
Languages
English
Size
10K–100K rows
Downloads this month
44K
Likes
184

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("eriktks/conll2003")