PuzzTE

PuzzTE数据集由克卢日-纳波卡理工大学计算机科学系的Roxana Szomiu和Adrian Groza创建，专注于自然语言推理任务。该数据集包含三种类型的逻辑谜题：比较谜题、骑士与无赖谜题和斑马谜题，总计756条。创建过程涉及使用自然语言处理技术自动识别谜题中的角色及其关系，并基于手动定义的语法规则将每个谜题转换为一阶逻辑（FOL）表示。数据集的应用领域包括机器理解任务，旨在通过逻辑谜题评估问答应用的能力。

The PuzzTE dataset was created by Roxana Szomiu and Adrian Groza from the Department of Computer Science, Politehnica University of Cluj-Napoca, and focuses on the natural language inference task. It contains three types of logical puzzles: comparison puzzles, knight-and-knave puzzles, and zebra puzzles, with a total of 756 entries. The dataset construction process involved using natural language processing technologies to automatically identify the characters and their relationships within the puzzles, and converting each puzzle into First-Order Logic (FOL) representation based on manually defined grammatical rules. The application scope of this dataset covers machine comprehension tasks, which aim to evaluate the capabilities of question-answering applications via logical puzzles.