Dataset Card for EBM NLP

This corpus release contains 4,993 abstracts annotated with (P)articipants, (I)nterventions, and (O)utcomes. Training labels are sourced from AMT workers and aggregated to reduce noise. Test labels are collected from medical professionals.

Citation Information

@inproceedings{nye-etal-2018-corpus,
    title = "A Corpus with Multi-Level Annotations of Patients, Interventions and Outcomes to Support Language Processing for Medical Literature",
    author = "Nye, Benjamin  and
      Li, Junyi Jessy  and
      Patel, Roma  and
      Yang, Yinfei  and
      Marshall, Iain  and
      Nenkova, Ani  and
      Wallace, Byron",
    booktitle = "Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2018",
    address = "Melbourne, Australia",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/P18-1019",
    doi = "10.18653/v1/P18-1019",
    pages = "197--207",
}