CICERO (Contextualized Commonsense Inference in Dialogues)

CICERO 包含从 5600 个对话中收集的五个常识维度（原因、后续事件、先决条件、动机和情绪反应）的 53,000 个推论。 IT 涉及两个具有挑战性的生成和多项选择任务，以供最先进的 NLP 模型解决。

CICERO contains 53,000 inferences across five commonsense dimensions (causes, subsequent events, preconditions, motivations, and emotional reactions) collected from 5,600 dialogues. It involves two challenging tasks, generation and multiple-choice questions, for state-of-the-art NLP models to solve.

搜集汇总

背景与挑战

背景概述

CICERO数据集包含从5600个对话中收集的五个常识维度的53,000个推论，用于支持生成和多项选择任务。该数据集旨在为自然语言处理模型提供上下文常识推理的挑战性测试。