CoNLL export
Campaigns with span questions export in CoNLL format: one token per line with a BIO tag, the format most NER training code reads.
What you get
Each sample becomes a block of lines, one token per line, with the token and a tab and a tag that marks where an entity begins (B-), continues (I-) or is outside any entity (O). The tags come from the spans a majority of reviewers marked.
Zarghona B-PERSON
visited O
Kandahar B-PLACE
in O
March OTokenisation
Spans are stored as character ranges, so tokens are split at whitespace for the export. Languages that do not separate words with spaces, or that attach clitics to words, may need your own tokeniser applied to the JSONL export instead.
Frequently asked questions
What is BIO tagging?
A scheme that labels each token as the Beginning of an entity, Inside an entity, or Outside any entity.
Can I get character offsets instead?
Yes. The JSONL and Label Studio exports keep the character start and end of every span.