relbert/scientific_and_creative_analogy

该数据集用于关系映射任务，即选择最佳的词对组合。关系映射`M`是在两组术语（`A`和`B`）之间的双射映射。给定术语集`A`和`B`，任务是识别映射`M`。数据实例包括源术语列表、目标术语列表和随机排列的目标术语列表，任务是从源到随机排列的目标找到正确的映射顺序。数据集包含45个测试实例。

This dataset is developed for the relation mapping task, which involves selecting the optimal word pair combination. The relation mapping `M` is a bijective mapping between two sets of terms, denoted as `A` and `B`. Given the term sets `A` and `B`, the task is to identify the mapping `M`. Each data instance includes a source term list, a target term list, and a randomly permuted target term list, with the task being to find the correct mapping order from the source terms to the randomly permuted target terms. The dataset contains 45 test instances.

原始信息汇总

数据集概述

基本信息

• 名称: Relation Mapping
• 语言: 英语
• 许可证: 其他
• 多语言性: 单语种
• 大小类别: 1<n<1K
• 美观名称: Relation Mapping

数据集描述

• 任务: 关系映射任务，旨在选择词对的最佳组合。
• 定义: 关系映射 M 是两个词集 (A 和 B) 之间的双射映射。
• 示例:
  • 集 A: ("solar system", "sun", "planet", "mass", "attracts", "revolves", "gravity")
  • 集 B: ("atom", "nucleus", "electron", "charge", "attracts", "revolves", "electromagnetism")
  • 映射 M:
    • "solar system" -> "atom"
    • "sun" -> "nucleus"
    • "planet" -> "electron"
    • "mass" -> "charge"
    • "attracts" -> "attracts"
    • "revolves" -> "revolves"
    • "gravity" -> "electromagnetism"

数据集结构

• 数据实例:
  • id: 实例标识
  • reference: 参考短语
  • source: 关系映射的源词列表
  • target: 与 source 对齐的目标词列表
  • target_random: 目标词的随机顺序列表
  • type: 实例类型（如：隐喻）
• 任务: 给定 source 和 target_random，预测 target_random 的正确顺序以匹配 target。
• 平均词数: 7个词
• 可能的顺序总数: 5040种

数据分割

• 分割名称: relation_mapping
• 测试集大小: 45个实例

引用信息

@article{czinczoll2022scientific, title={Scientific and Creative Analogies in Pretrained Language Models}, author={Czinczoll, Tamara and Yannakoudakis, Helen and Mishra, Pushkar and Shutova, Ekaterina}, journal={arXiv preprint arXiv:2211.15268}, year={2022} }