Kicaulah/script_daataset_ai
收藏资源简介:
EcommerceAI数据集是一个高质量的合成英文电子商务客户服务对话数据集,包含完整的意图和情感标注。该数据集由5,000个对话组成,总计47,028个对话轮次,平均每个对话9.4轮。数据覆盖5种问题类别:延迟交付、收到错误商品、退款请求、商品损坏和产品咨询,以及10种产品类别和12种意图类别。所有对话均为程序生成,不包含真实客户数据,每个记录在元数据中标记为SYNTHETIC。数据集适用于微调LLMs用于客户服务聊天机器人、训练意图分类器和情感分析模型、对话状态跟踪研究等任务。
The EcommerceAI Dataset is a high-quality synthetic English-language dataset of e-commerce customer service conversations with full intent and sentiment annotations. It consists of 5,000 conversations, totaling 47,028 dialogue turns, with an average of 9.4 turns per conversation. The dataset covers 5 issue categories: Late Delivery, Wrong Item Received, Refund Request, Damaged Item, and Product Inquiry, along with 10 product categories and 12 intent classes. All conversations are programmatically generated, and no real customer data is included, with each record labeled as SYNTHETIC in metadata. It is suitable for fine-tuning LLMs for customer service chatbots, training intent classifiers and sentiment analysis models, dialogue state tracking research, and more.




