Logic2Text
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/czyssrs/logic2text
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为Logic2Text，包含了从结构化逻辑形式到自然语言生成的任务，分为训练样本、验证样本和测试样本三个部分。其中，自然语言段的最大长度为130个单词，超过90%的条目少于90个单词。该数据集的规模为：训练样本7566个，验证样本1000个，测试样本1095个，其任务是逻辑形式到自然语言的生成。
The dataset is named Logic2Text, which focuses on the task of generating natural language from structured logical forms, and is divided into three subsets: training, validation, and test samples. The maximum length of the natural language segments is 130 words, and over 90% of the entries contain fewer than 90 words. The dataset has 7566 training samples, 1000 validation samples, and 1095 test samples, with its core task being natural language generation from structured logical forms.
