Datasets:

araag2
/

SemEval_NLI4CT

dataset_info:
  - config_name: 2023_conversational
    features:
      - name: id
        dtype: string
      - name: prompt
        list:
          - name: role
            dtype: string
          - name: content
            dtype: string
      - name: completion
        list:
          - name: role
            dtype: string
          - name: content
            dtype: string
      - name: Label
        dtype: string
    splits:
      - name: train
        num_bytes: 4686231
        num_examples: 1700
      - name: dev
        num_bytes: 522955
        num_examples: 200
      - name: test
        num_bytes: 1465772
        num_examples: 500
    download_size: 2386772
    dataset_size: 6674958
  - config_name: 2023_processed
    features:
      - name: id
        dtype: string
      - name: Type
        dtype: string
      - name: Section_id
        dtype: string
      - name: Primary_id
        dtype: string
      - name: Secondary_id
        dtype: string
      - name: Statement
        dtype: string
      - name: Label
        dtype: string
      - name: Primary_evidence_index
        sequence: int64
      - name: Secondary_evidence_index
        sequence: int64
      - name: Primary_Document
        struct:
          - name: Adverse Events
            sequence: string
          - name: Eligibility
            sequence: string
          - name: Intervention
            sequence: string
          - name: Results
            sequence: string
      - name: Primary_Section
        sequence: string
      - name: Secondary_Document
        struct:
          - name: Adverse Events
            sequence: string
          - name: Eligibility
            sequence: string
          - name: Intervention
            sequence: string
          - name: Results
            sequence: string
      - name: Secondary_Section
        sequence: string
    splits:
      - name: train
        num_bytes: 15485945
        num_examples: 1700
      - name: dev
        num_bytes: 1761256
        num_examples: 200
      - name: test
        num_bytes: 4873788
        num_examples: 500
    download_size: 7268052
    dataset_size: 22120989
  - config_name: 2023_source
    features:
      - name: id
        dtype: string
      - name: Type
        dtype: string
      - name: Section_id
        dtype: string
      - name: Primary_id
        dtype: string
      - name: Secondary_id
        dtype: string
      - name: Statement
        dtype: string
      - name: Label
        dtype: string
      - name: Primary_evidence_index
        sequence: int64
      - name: Secondary_evidence_index
        sequence: int64
    splits:
      - name: train
        num_bytes: 571852
        num_examples: 1700
      - name: dev
        num_bytes: 61121
        num_examples: 200
      - name: test
        num_bytes: 175674
        num_examples: 500
    download_size: 283972
    dataset_size: 808647
  - config_name: 2024_conversational
    features:
      - name: id
        dtype: string
      - name: prompt
        list:
          - name: role
            dtype: string
          - name: content
            dtype: string
      - name: completion
        list:
          - name: role
            dtype: string
          - name: content
            dtype: string
      - name: Label
        dtype: string
    splits:
      - name: test
        num_bytes: 16283587
        num_examples: 5500
      - name: dev
        num_bytes: 522955
        num_examples: 200
      - name: train
        num_bytes: 4686231
        num_examples: 1700
    download_size: 7652506
    dataset_size: 21492773
  - config_name: 2024_processed
    features:
      - name: id
        dtype: string
      - name: Type
        dtype: string
      - name: Section_id
        dtype: string
      - name: Primary_id
        dtype: string
      - name: Statement
        dtype: string
      - name: Label
        dtype: string
      - name: Intervention
        dtype: string
      - name: Causal_type
        sequence: string
      - name: Secondary_id
        dtype: string
      - name: Secondary_evidence_index
        dtype: 'null'
      - name: Primary_Document
        struct:
          - name: Adverse Events
            sequence: string
          - name: Eligibility
            sequence: string
          - name: Intervention
            sequence: string
          - name: Results
            sequence: string
      - name: Primary_Section
        sequence: string
      - name: Secondary_Document
        struct:
          - name: Adverse Events
            sequence: string
          - name: Eligibility
            sequence: string
          - name: Intervention
            sequence: string
          - name: Results
            sequence: string
      - name: Secondary_Section
        sequence: string
    splits:
      - name: test
        num_bytes: 53427368
        num_examples: 5500
      - name: dev
        num_bytes: 1748808
        num_examples: 200
      - name: train
        num_bytes: 15342665
        num_examples: 1700
    download_size: 13749622
    dataset_size: 70518841
  - config_name: 2024_source
    features:
      - name: id
        dtype: string
      - name: Type
        dtype: string
      - name: Section_id
        dtype: string
      - name: Primary_id
        dtype: string
      - name: Statement
        dtype: string
      - name: Label
        dtype: string
      - name: Intervention
        dtype: string
      - name: Causal_type
        sequence: string
      - name: Secondary_id
        dtype: string
      - name: Secondary_evidence_index
        dtype: 'null'
    splits:
      - name: test
        num_bytes: 1965546
        num_examples: 5500
      - name: dev
        num_bytes: 48673
        num_examples: 200
      - name: train
        num_bytes: 428572
        num_examples: 1700
    download_size: 1034926
    dataset_size: 2442791
configs:
  - config_name: 2023_conversational
    data_files:
      - split: train
        path: 2023_conversational/train-*
      - split: dev
        path: 2023_conversational/dev-*
      - split: test
        path: 2023_conversational/test-*
  - config_name: 2023_processed
    data_files:
      - split: train
        path: 2023_processed/train-*
      - split: dev
        path: 2023_processed/dev-*
      - split: test
        path: 2023_processed/test-*
  - config_name: 2023_source
    data_files:
      - split: train
        path: 2023_source/train-*
      - split: dev
        path: 2023_source/dev-*
      - split: test
        path: 2023_source/test-*
  - config_name: 2024_conversational
    data_files:
      - split: test
        path: 2024_conversational/test-*
      - split: dev
        path: 2024_conversational/dev-*
      - split: train
        path: 2024_conversational/train-*
  - config_name: 2024_processed
    data_files:
      - split: test
        path: 2024_processed/test-*
      - split: dev
        path: 2024_processed/dev-*
      - split: train
        path: 2024_processed/train-*
  - config_name: 2024_source
    data_files:
      - split: test
        path: 2024_source/test-*
      - split: dev
        path: 2024_source/dev-*
      - split: train
        path: 2024_source/train-*
license: cc-by-sa-4.0
task_categories:
  - text-classification
  - question-answering
language:
  - en
tags:
  - medical
pretty_name: 'NLI4CT: Natural Language Inference for Clinical Trial Reports'
size_categories:
  - 1K<n<10K

NLI4CT: Multi-Evidence Natural Language Inference for Clinical Trial Reports and SemEval-2024 Task 2: Safe Biomedical Natural Language Inference for Clinical Trials

Dataset Description

	Links
Homepage:	sites.google
Repository:	Github2024
Paper:	arXiv2023 / arXiv2024
Leaderboard:	Codalab2023
Contact (Original Authors):	Maël Jullien (mael.jullien@postgrad.manchester.ac.uk)
Contact (Curator):	Artur Guimarães (artur.guimas@gmail.com)

Dataset Summary

The NLI4CT dataset introduces a challenging two-part benchmark designed to enable large-scale automated reasoning over full clinical trial reports (CTRs): (1) determining whether a natural language statement is entailed or contradicted by a CTR (textual entailment) and (2) retrieving the specific evidence sentences that justify that label. It covers 2,400 expert-annotated instances drawn from breast cancer trials, each mapped to one of four CTR sections—eligibility, intervention, results, or adverse events—and includes both single-trial and comparison scenarios.

Data Instances

Source Format

{
        "Type": "Comparison",
        "Section_id": "Eligibility",
        "Primary_id": "NCT01129622",
        "Secondary_id": "NCT01156987",
        "Statement": "Women suffering from both claustrophobia and IBS or not eligible for either the primary trial or the secondary trial.",
        "Label": "Contradiction",
        "Primary_evidence_index": [
            2,
            3
        ],
        "Secondary_evidence_index": [
            2,
            9
        ]
}

Data Fields

Source Format

TO:DO

Data Splits

TO:DO

Additional Information

Dataset Curators

Original Paper

Maël Jullien - Department of Computer Science, University of Manchester, United Kingdom
Marco Valentino - Idiap Research Institute, Switzerland
Hannah Frost - Department of Computer Science, University of Manchester, United Kingdom, and Digital Experimental Cancer Medicine Team, Cancer Research UK Manchester Institute
Paul O’Regan - Digital Experimental Cancer Medicine Team, Cancer Research UK Manchester Institute
Donal Landers- Digital Experimental Cancer Medicine Team, Cancer Research UK Manchester Institute
André Freitas - Department of Computer Science, University of Manchester, United Kingdom and Digital Experimental Cancer Medicine Team, Cancer Research UK Manchester Institute and Idiap Research Institute, Switzerland

Huggingface Curator

Artur Guimarães (artur.guimas@gmail.com) - INESC-ID / University of Lisbon - Instituto Superior Técnico

Licensing Information

CC BY-SA 4.0

Citation Information

@article{jullien2023semeval,
  title={SemEval-2023 task 7: Multi-evidence natural language inference for clinical trial data},
  author={Jullien, Ma{\"e}l and Valentino, Marco and Frost, Hannah and O'Regan, Paul and Landers, Donal and Freitas, Andr{\'e}},
  journal={arXiv preprint arXiv:2305.02993},
  year={2023}
}

@article{jullien2024semeval,
  title={SemEval-2024 task 2: Safe biomedical natural language inference for clinical trials},
  author={Jullien, Ma{\"e}l and Valentino, Marco and Freitas, Andr{\'e}},
  journal={arXiv preprint arXiv:2404.04963},
  year={2024}
}

10.48550/ARXIV.2305.02993 10.48550/ARXIV.2404.04963

Contributions

Thanks to araag2 for adding this dataset.