Ayushnangia/SynWOZ

SynWOZ数据集包含50k个对话，这些对话通过先进的对话生成管道生成，模拟了餐厅、酒店、出租车等多种服务中的真实交互。对话涵盖了多种场景、情感和解决状态。数据集主要用于对话建模、情感识别、意图分类和对话AI研究。数据集的结构包括服务列表、对话ID、对话轮次、用户和助理的情感、场景类别、生成场景、时间段、地理区域和解决状态等字段。数据集是单语种（英语），并且没有预定义的数据分割。数据集的生成过程涉及人物管理、场景生成、对话生成、唯一性验证和情感分配。数据集的来源数据包括multi_woz_v22数据集和FinePersonas-v0.1-clustering-100k数据集。数据集的注释是机器生成的，注释字段包括意图、情感和场景细节。

The SynWOZ dataset contains 50k dialogues generated using an advanced dialogue generation pipeline, simulating realistic interactions across various services such as restaurants, hotels, taxis, and more. The dialogues incorporate diverse scenarios, emotions, and resolution statuses. The dataset is primarily used for dialogue modeling, emotion recognition, intent classification, and conversational AI research. The dataset structure includes fields such as services, dialogue ID, turns, user and assistant emotions, scenario category, generated scenario, time slot, regions, and resolution status. The dataset is monolingual (English) and does not have predefined splits. The dataset creation process involves persona management, scenario generation, dialogue generation, uniqueness verification, and emotion assignment. The source data includes the multi_woz_v22 dataset and the FinePersonas-v0.1-clustering-100k dataset. The annotations are machine-generated, with fields including intents, emotions, and scenario details.