neil-code/dialogsum-test
收藏资源简介:
DIALOGSum是一个大规模的对话摘要数据集,包含13,460个对话(外加100个用于主题生成的保留数据)及其对应的人工标注的摘要和主题。数据集来源于多个公开的对话语料库,涵盖了日常生活中的多种场景,如学校、工作、医疗、购物、休闲、旅行等。对话通常发生在朋友、同事以及服务提供者与顾客之间。与之前的对话数据集相比,DIALOGSum的对话具有更丰富的现实生活场景、更清晰的沟通模式和意图,以及适合自动摘要的合理长度。每个对话的摘要由语言专家根据特定标准进行标注,包括传达最重要的信息、简洁、保留对话中的重要命名实体、以观察者视角书写、使用正式语言等。
DIALOGSum is a large-scale dialogue summarization dataset containing 13,460 dialogues (plus 100 reserved samples for topic generation) along with their manually annotated summaries and topics. The dataset is sourced from multiple public dialogue corpora, covering various daily life scenarios such as school, work, healthcare, shopping, leisure, travel, and more. Dialogues typically occur between friends, colleagues, as well as service providers and customers. Compared with previous dialogue datasets, the dialogues in DIALOGSum feature more diverse real-life scenarios, clearer communication patterns and intentions, as well as reasonable lengths suitable for automatic summarization. The summary for each dialogue is annotated by language experts in accordance with specific criteria, including conveying the most important information, being concise, retaining important named entities in the dialogue, being written from an observer's perspective, and using formal language, among others.
DIALOGSum Corpus 数据集概述
数据集描述
数据集摘要
DialogSum 是一个大规模的对话摘要数据集,包含 13,460 个对话及其对应的手动标注摘要和主题。此外,还有 100 个用于主题生成的保留数据。
语言
英语
数据集结构
数据实例
DialogSum 数据集包含 13,460 个对话(外加 1000 个测试数据),分为训练集、测试集和验证集。
训练集中的第一个实例: json { "id": "train_0", "summary": "Mr. Smiths getting a check-up, and Doctor Hawkins advises him to have one every year. Hawkinsll give some information about their classes and medications to help Mr. Smith quit smoking.", "dialogue": "#Person1#: Hi, Mr. Smith. Im Doctor Hawkins. Why are you here today? #Person2#: I found it would be a good idea to get a check-up. #Person1#: Yes, well, you havent had one for 5 years. You should have one every year. #Person2#: I know. I figure as long as there is nothing wrong, why go see the doctor? #Person1#: Well, the best way to avoid serious illnesses is to find out about them early. So try to come at least once a year for your own good. #Person2#: Ok. #Person1#: Let me see here. Your eyes and ears look fine. Take a deep breath, please. Do you smoke, Mr. Smith? #Person2#: Yes. #Person1#: Smoking is the leading cause of lung cancer and heart disease, you know. You really should quit. #Person2#: Ive tried hundreds of times, but I just cant seem to kick the habit. #Person1#: Well, we have classes and some medications that might help. Ill give you more information before you leave. #Person2#: Ok, thanks doctor.", "topic": "get a check-up" }
数据字段
dialogue: 对话文本。summary: 人工编写的对话摘要。topic: 人工编写的对话主题/一句话概括。id: 示例的唯一文件ID。
数据分割
- 训练集: 12460
- 验证集: 500
- 测试集: 1500
- 保留集: 100(仅包含
id,dialogue,topic三个字段)
数据集创建
策划理由
DialogSum 数据集从三个公共对话语料库(Dailydialog、DREAM 和 MuTual)以及一个英语口语练习网站收集对话数据。这些数据集包含面对面口语对话,涵盖日常生活主题,如学校、工作、医疗、购物、休闲、旅行等。大多数对话发生在朋友、同事以及服务提供者和顾客之间。
与先前的数据集相比,DialogSum 的对话具有以下特点:
- 在丰富的真实生活场景中,包括更多样化的任务导向场景;
- 具有清晰的沟通模式和意图,适合作为摘要的来源;
- 具有合理的长度,适合自动摘要的目的。
标注者根据以下标准对每个对话进行摘要:
- 传达最重要的信息;
- 简洁;
- 保留对话中的重要命名实体;
- 从观察者角度编写;
- 使用正式语言。
源语言生产者
语言学家
标注者
语言专家
许可信息
MIT 许可证




