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", }