数据集:
bigbio/ebm_pico
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.
@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",
}