Topical-Chat
收藏资源简介:
我们介绍了 Topical-Chat,这是一个以知识为基础的人与人对话数据集,其中基础知识跨越 8 个广泛的主题,对话伙伴没有明确定义的角色。 Topical-Chat 大致由两种类型的文件组成: (1) 对话文件 - 这些是 .json 文件,其中包含 Amazon Mechanical Turk(也称为 Turkers)上的两名工人之间的对话 (2) 阅读集文件 - 这些是 .json 文件,其中包含来自不同数据源(维基百科、Reddit 和华盛顿邮报)的知识部分,供特定的 Turker 在对话时阅读和参考。
We introduce Topical-Chat, a knowledge-grounded human-to-human dialogue dataset. The underlying knowledge spans 8 broad topics, and the dialogue partners have no explicitly defined roles. Topical-Chat roughly consists of two types of files: (1) Dialogue files: These are .json files containing conversations between two workers on Amazon Mechanical Turk (also referred to as Turkers). (2) Reading set files: These are .json files containing knowledge snippets from diverse data sources including Wikipedia, Reddit, and The Washington Post, for specific Turkers to read and reference during their dialogues.

- Topical-Chat数据集首次发表,由Karthik Gopalakrishnan等人提出,旨在促进对话系统在特定主题上的深入交流。
- Topical-Chat数据集首次应用于对话生成模型的训练,展示了其在提升对话质量和深度方面的潜力。
- Topical-Chat数据集被广泛用于多轮对话系统的研究,成为评估对话模型性能的重要基准之一。
- 1Topical-Chat: A Manually-Collected Corpus for Knowledge-Grounded ConversationUniversity of Michigan, University of Washington · 2019年
- 2Improving Knowledge-Grounded Dialogue Systems with Compositional Knowledge TransformersUniversity of California, Berkeley · 2021年
- 3Knowledge-Grounded Dialogue Generation with Pre-trained Language ModelsTsinghua University, Microsoft Research · 2020年
- 4Evaluating the Impact of Knowledge Graph Structure on Knowledge-Grounded Dialogue GenerationUniversity of Cambridge · 2021年
- 5A Survey on Knowledge-Enhanced Dialogue SystemsTsinghua University, University of Cambridge · 2021年



