CINLID

CINLID数据集包含106,000对人工生成的中文成语，是一个语义推理数据集。该数据集包含少量简洁文本，如谜语和口语表达。每一对成语分别作为前提和假设，用于评估它们之间的语义关系，这些关系可以是近义、无关或对立。

The CINLID dataset is a semantic reasoning dataset containing 106,000 manually generated pairs of Chinese idioms. It also includes a small quantity of concise textual samples such as riddles and colloquial expressions. Each pair of idioms serves as the premise and hypothesis respectively, which is used to evaluate the semantic relationships between them, covering near-synonymous, irrelevant, and contradictory relations.

搜集汇总

背景与挑战

背景概述

CINLID是一个包含106832条人工标注中文成语对的数据集，支持自然语言推理任务，标注标签为entailment、contradiction和neutral，旨在帮助机器理解成语的语义关系。