W-L/Customer-service-tickets-qwen-qa

--- license: cc-by-nc-4.0 pretty_name: Customer Support Tickets QA en SFT size_categories: - 10K<n<100K task_categories: - text-generation - question-answering - text-classification language: - en tags: - customer-support - helpdesk - tickets - instruction-tuning - sft - qwen configs: - config_name: default data_files: - split: train path: customer_support_tickets_en_qwen3_5.jsonl --- # Customer Support Tickets QA (English) — Qwen SFT Dataset This dataset is formatted for **supervised fine-tuning (SFT)** of Qwen-style chat models on customer support email tasks. source dataset: Tobi-Bueck/customer-support-tickets It is designed for training models to read a customer ticket, understand its context, and generate an appropriate support response. Depending on the prompt design, the same data can also support auxiliary tasks such as queue prediction, priority prediction, and ticket type classification. ## Dataset Format The training file is: - `customer_support_tickets_en_qwen3_5.jsonl` Each row is a single SFT example in a **Qwen chat template–compatible structure**. Typical source fields described by the original dataset include: - **queue**: department responsible for handling the ticket - **priority**: urgency level of the issue - **language**: language of the ticket - **subject**: subject line of the customer email - **body**: full customer message - **answer**: support agent response - **type**: ticket type, such as Incident, Request, Problem, or Change - **business_type**: domain of the helpdesk or company - **tags**: issue labels or categories These fields are reflected in the prompt so the model can learn both response generation and structured reasoning over ticket metadata. ## Intended Use This dataset is intended for: - **Instruction tuning / supervised fine-tuning** - **Customer support response generation** - **Ticket understanding** - **Synthetic helpdesk assistant training**许可协议：知识共享署名-非商业性使用4.0（CC BY-NC 4.0） 规范名称：客户支持工单问答英文监督微调数据集 规模分类：10,000 < 样本量 < 100,000 任务类别：文本生成、问答、文本分类 语言：英语 标签：客户支持、服务台、工单、指令微调、监督微调（SFT）、通义千问（Qwen） 配置项： - 配置名称：默认 数据文件： - 拆分方式：训练集 路径：customer_support_tickets_en_qwen3_5.jsonl # 英文客户支持工单问答——通义千问监督微调数据集 本数据集专为针对客户支持邮件任务的**监督微调（SFT）**训练通义千问风格对话模型而设计，其源数据集为Tobi-Bueck/customer-support-tickets。 本数据集旨在训练模型读取客户工单、理解上下文并生成恰当的支持回复。根据提示词设计的不同，该数据集亦可支撑队列预测、优先级预测、工单类型分类等辅助任务。 ## 数据集格式 训练文件为： - `customer_support_tickets_en_qwen3_5.jsonl` 每一行均为符合**通义千问对话模板兼容结构**的单条监督微调样本。 源数据集的典型字段如下： - **服务队列（queue）**：负责处理该工单的部门 - **优先级（priority）**：问题的紧急程度 - **语言（language）**：工单的语言 - **主题（subject）**：客户邮件的主题行 - **正文（body）**：客户的完整留言 - **回复（answer）**：客服专员的回复 - **工单类型（type）**：工单类别，如事件（Incident）、请求（Request）、问题（Problem）或变更（Change） - **业务类型（business_type）**：服务台或所属企业的业务领域 - **标签（tags）**：问题的标注或分类 提示词中涵盖了上述字段，因此模型可同时学习回复生成与工单元数据的结构化推理能力。 ## 预期用途 本数据集适用于： - **指令微调/监督微调** - **客户支持回复生成** - **工单理解** - **虚拟服务台助手训练**