sergioburdisso/dialog2flow-dataset

Dialog2Flow Training Corpus数据集包含340万条带有对话行为和槽位标签的句子，涵盖了52个不同的领域。该数据集来源于20个标准化的任务导向对话数据集，并提供了三种训练数据集：包含对话行为和槽位标签的D2F_single数据集，以及分别包含对话行为和槽位标签的DFD_joint数据集。此外，数据集还提供了20个原始数据集的详细信息，包括每个数据集的名称、训练集、验证集、测试集的大小以及许可证信息。数据集的统计信息包括对话行为、槽位和领域的详细列表。

The Dialog2Flow Training Corpus dataset consists of 3.4 million utterances/sentences annotated with dialog act and slot labels across 52 different domains. The dataset is built from 20 standardized task-oriented dialogue datasets and provides three training datasets: D2F_single containing both dialog act and slot annotations, and DFD_joint containing subsets with dialog acts and slots separately. Additionally, the dataset provides detailed information on the 20 original datasets, including the name, size of the training, validation, and test sets, and license information for each dataset. The datasets statistics include detailed lists of dialog acts, slots, and domains.