CoXQL

CoXQL是由德国人工智能研究中心创建的第一个用于解析对话式可解释AI系统中解释请求的数据集。该数据集包含1179对用户问题和相应的SQL-like查询，专门设计用于增强可解释AI系统的意图识别能力。数据集的创建过程涉及从现有系统中选择和手动创建问题与解析对，并使用ChatGPT进行扩充。CoXQL旨在通过提供丰富的意图识别训练数据，解决当前可解释AI系统在自由对话中识别用户意图的挑战，特别是在处理具有多个槽位的复杂操作时。该数据集的应用领域包括提升对话式可解释AI系统的用户交互体验和信任度，以及作为未来研究的基准。

CoXQL is the first dataset dedicated to parsing explanation requests in conversational explainable AI (XAI) systems, developed by the German Research Center for Artificial Intelligence (DFKI). This dataset comprises 1,179 pairs of user questions and their corresponding SQL-like queries, and is specifically engineered to bolster the intent recognition capabilities of explainable AI systems. The development of CoXQL involved selecting and manually crafting question-parsing pairs from existing systems, followed by data augmentation using ChatGPT. CoXQL aims to address the challenge of user intent recognition for contemporary explainable AI systems in open-domain conversational scenarios, particularly when handling complex operations involving multiple slots, by providing rich training data for intent recognition tasks. The application scenarios of this dataset include enhancing user interaction experience and trust in conversational explainable AI systems, as well as serving as a benchmark for future research.