worta/apty

APTY数据集来自论文《Towards Human Understanding of Paraphrase Types in ChatGPT》，包含两个部分：APTY_base和APTY_ranked。APTY_base包含基于ETPC数据集的特定原子释义类型的注释释义，而APTY_ranked包含人类对特定原子释义类型的释义偏好排名。生成释义候选的代码和原始数据可以在GitHub上找到。

The APTY dataset is from the paper Towards Human Understanding of Paraphrase Types in ChatGPT. It consists of two parts: The first part (APTY_base) contains annotated paraphrases with specific atomic paraphrase types based on the ETPC dataset. The second part (APTY_ranked) consists of human preferences ranking paraphrases with specific atomic paraphrase types. The code to generate the paraphrase candidates can be found on GitHub and uses ChatGPT. The raw data is also available on GitHub.

原始信息汇总

APTY 数据集

概述

• 名称: APTY
• 来源: 论文 "Towards Human Understanding of Paraphrase Types in ChatGPT"
• 许可证: CC BY-SA 4.0

数据集结构

• APTY-base:

  • 文件: apty_base.parquet
  • 内容: 包含基于 ETPC 数据集的注释释义，具有特定的原子释义类型。

• APTY-ranked:

  • 文件: apty_ranked_hg.parquet
  • 内容: 包含人类偏好排序的释义，具有特定的原子释义类型。

相关资源

• 释义生成代码: GitHub 仓库
• 原始数据: GitHub 仓库

引用

bib @misc{meier2024humanunderstandingparaphrasetypes, title={Towards Human Understanding of Paraphrase Types in ChatGPT}, author={Dominik Meier and Jan Philip Wahle and Terry Ruas and Bela Gipp}, year={2024}, eprint={2407.02302}, archivePrefix={arXiv}, primaryClass={cs.CL}, url={https://arxiv.org/abs/2407.02302}, }