DebateLabKIT/deepa2
收藏Hugging Face2024-11-23 更新2024-03-04 收录
下载链接：
https://hf-mirror.com/datasets/DebateLabKIT/deepa2
下载链接
链接失效反馈官方服务：
资源简介：
这是一个不断增长的、经过精心策划的`deepa2`数据集集合，包含了对论证性文本的全面逻辑分析。该集合包括：通过[`deepa2 bake`](https://github.com/debatelab/deepa2)工具从现有NLP数据集构建的数据集，以及专门为该集合创建的原始`deepa2`数据集。该数据集支持的任务包括条件文本生成、结构预测和文本检索。数据集的语言为英语，未来将扩展到其他语言。数据集结构包括多个子数据集，每个子数据集包含训练、验证和测试三个分割。数据集的创建目的是统一逻辑分析和论证重建相关的NLP资源。
This is an expanding, carefully curated collection of the `deepa2` datasets, which encompasses comprehensive logical analyses of argumentative texts. This collection includes datasets constructed from existing NLP datasets via the [`deepa2 bake`](https://github.com/debatelab/deepa2) tool, as well as original `deepa2` datasets created exclusively for this collection. The tasks supported by these datasets include conditional text generation, structure prediction, and text retrieval. The datasets are currently in English, with plans to expand to other languages in the future. The dataset structure consists of multiple sub-datasets, each of which contains three splits: training, validation, and test. The purpose of creating this collection is to unify NLP resources related to logical analysis and argument reconstruction.
提供机构：
DebateLabKIT原始信息汇总
数据集概述
数据集名称
deepa2 Datasets Collection
语言
- 主要语言：英语
- 未来计划扩展至其他语言
许可证
- 许可证类型：其他
多语言性
- 单语种
大小分类
- 未知
任务类别
- 文本检索
- 文本生成
任务ID
- 文本简化
- 解析
标签
- 论证挖掘
- 摘要
- 条件文本生成
- 结构预测
数据集结构
子数据集
esnli: 从e-SNLI创建enbank: 从Entailment Bank创建argq: 从IBM-ArgQ创建argkp: 从IBM-KPA创建aifdb: 从AIFdb创建aaac: 原始数据集，机器生成
数据字段
| 特征 | esnli | enbank | aifdb | aaac | argq | argkp |
|---|---|---|---|---|---|---|
source_text |
x | x | x | x | x | x |
title |
x | x | ||||
gist |
x | x | x | x | ||
source_paraphrase |
x | x | x | x | ||
context |
x | x | x | |||
reasons |
x | x | x | x | x | |
conjectures |
x | x | x | x | x | |
argdown_reconstruction |
x | x | x | |||
erroneous_argdown |
x | x | ||||
premises |
x | x | x | x | x | |
intermediary_conclusion |
x | |||||
conclusion |
x | x | x | x | x | |
premises_formalized |
x | x | x | x | ||
intermediary_conclusion_formalized |
x | |||||
conclusion_formalized |
x | x | x | x | ||
predicate_placeholders |
x | |||||
entity_placeholders |
x | |||||
misc_placeholders |
x | x | x | |||
plchd_substitutions |
x | x | x |
数据分割
- 每个子数据集包含三个分割：
train,validation,test
许可证信息
| 子数据集 | 许可证 |
|---|---|
| esnli | MIT |
| aifdb | 学术免费使用 |
| enbank | CC BY 4.0 |
| aaac | CC BY 4.0 |
| argq | CC BY SA 4.0 |
| argkp | Apache |
