TaxiNLI

TaxiNLI 是根据上述分类的原则和分类收集的数据集。通过基于蕴涵标签和域进行均匀采样，从 MultiNLI（Williams 等人，2018 年）中精选了一个示例子集。数据集使用细粒度的类别标签进行注释。TaxiNLI is a dataset collected in accordance with the aforementioned classification principles and categories. A representative subset was selected from MultiNLI (Williams et al., 2018) via uniform sampling based on entailment labels and domains. The dataset is annotated with fine-grained category labels.

搜集汇总

背景与挑战

背景概述

TaxiNLI 是一个自然语言推理数据集，通过从MultiNLI中均匀采样构建，包含细粒度的类别标签。该数据集由Google Research、微软印度研究院和里士满大学于2020年联合发布，相关资源已开源。