The file claims_with_citances.jsonl contains data to allow researchers to train "claim generation" models: given a citation context, generate self-contained "atomic" claims based on the context.
Each line in the file contains information on the claims written based on a single citaion sentence. The fiels are:
s2orc_id: The S2ORC document ID of the document containing the citance.title: The title of the paper.abstract: The abstract of the paper.citation_paragraph: The paragraph containing the citation sentence.citance: The citation sentence on which the claim is based.claims: A list of claims, written based on the citance. Each claim has the following fields:claim_text: The text of the claim.claim_subject: The subject of the claim. May be blank.is_negation: Does this claim negate the meaning of the citance?