Sentences Involving Complex Compositional Knowledge (SICCK)

Sentences Involving Complex Compositional Knowledge (SICCK) 数据集由亚利桑那大学创建，包含1304对句子，用于评估自然语言推理（NLI）模型在理解逻辑组合性方面的表现。该数据集通过对SICK数据集中的15个示例进行修改而生成，使用了包括全称量词、存在量词、否定和其他自然逻辑概念修饰词在内的一系列修饰词来修改前提和假设。数据集的创建过程涉及对原始文本的修改、语义分析和根据自然逻辑规则重新标注蕴涵标签。SICCK数据集主要应用于自然语言处理领域，旨在解决NLI模型在处理组合性逻辑时的性能问题。

Sentences Involving Complex Compositional Knowledge (SICCK) dataset was created by the University of Arizona, which contains 1,304 sentence pairs for evaluating the performance of natural language inference (NLI) models in understanding logical compositionality. This dataset is generated by modifying 15 examples from the original SICK dataset, using a series of modifiers including universal quantifiers, existential quantifiers, negation, and other natural logic concept modifiers to revise both the premise and hypothesis. The dataset creation process involves original text modification, semantic analysis, and re-annotation of entailment labels according to natural logic rules. The SICCK dataset is primarily applied in the field of natural language processing, aiming to address the performance issues of NLI models when handling compositional logic.