Action-Based Conversations Dataset (ABCD)
收藏资源简介:
Action-Based Conversations Dataset (ABCD) 是一个大规模的标注数据集,由ASAPP和哥伦比亚大学创建。该数据集包含10042个对话,涉及55种用户意图,每个意图都需要遵循特定的行动序列来解决。ABCD数据集的特点在于其对话内容需要代理同时处理用户请求和内部政策限制,这使得数据集在模拟真实客户服务场景中更为逼真。数据集的应用领域主要集中在提高任务导向对话系统的深度和复杂性,特别是在需要遵循多步骤程序和公司政策的客户支持交互中。
Action-Based Conversations Dataset (ABCD) is a large-scale annotated dataset co-developed by ASAPP and Columbia University. It comprises 10,042 conversations spanning 55 user intents, each requiring a specific action sequence for resolution. A key characteristic of the ABCD dataset is that its conversations require agents to concurrently handle both user requests and internal policy constraints, rendering it more realistic for simulating real-world customer service scenarios. The primary application domains of this dataset focus on enhancing the depth and complexity of task-oriented dialogue systems, especially in customer support interactions that demand adherence to multi-step procedures and corporate policies.




