An Evaluation Dataset for Intent Classification and Out-of-Scope Prediction

本数据集由Clinc, Inc.创建，包含23,700条查询，涵盖150个意图类别及超出系统支持范围的查询。数据集通过众包方式收集，包括22,500条内范围查询和1,200条超出范围查询，涉及10个通用领域。创建过程包括定义意图、收集和预处理数据。该数据集主要用于评估任务驱动对话系统中的意图分类和超出范围预测，旨在提高系统对未知查询的处理能力。

This dataset was created by Clinc, Inc., containing 23,700 queries spanning 150 intent categories and out-of-scope (OOS) queries that are beyond the system's supported scope. It was collected via crowdsourcing, consisting of 22,500 in-scope queries and 1,200 OOS queries across 10 general domains. The dataset creation process includes defining intents, data collection and preprocessing. This dataset is primarily used for evaluating intent classification and out-of-scope prediction in task-oriented dialogue systems, aiming to improve the system's capability to handle unknown user queries.