哈利波特对话(HPD)数据集
收藏资源简介:
哈利波特对话(HPD)数据集是由腾讯AI实验室和香港科技大学(广州)联合创建的双语数据集,旨在推进对话代理与故事角色对齐的研究。该数据集包含《哈利波特》系列小说中的所有对话,涵盖英语和中文版本,并附有重要的背景信息,如对话场景、说话者、角色关系和属性。HPD数据集不仅为大型语言模型提供了丰富的角色驱动对话能力,还可用作评估语言模型与特定角色对齐程度的通用基准。此外,数据集的构建过程中,特别关注了场景、属性和关系的动态变化,以更准确地反映故事情节的发展。HPD数据集的应用领域广泛,特别是在提升对话代理与故事角色的对齐度方面,展示了其重要价值。
The Harry Potter Dialogue (HPD) dataset is a bilingual dataset jointly created by Tencent AI Lab and The Hong Kong University of Science and Technology (Guangzhou), aiming to advance research on aligning dialogue agents with fictional story characters. This dataset covers all dialogues from the Harry Potter novel series, including both English and Chinese versions, and is supplemented with critical background information such as dialogue scenes, speakers, character relationships and attributes. The HPD dataset not only provides rich role-driven dialogue capabilities for large language models (LLMs), but also serves as a universal benchmark for evaluating the alignment degree between language models and specific fictional characters. Furthermore, during the dataset construction process, special attention was paid to the dynamic changes of scenes, attributes and relationships to more accurately reflect the progression of the story plot. The HPD dataset has a wide range of application scenarios, and its important value is particularly highlighted in improving the alignment between dialogue agents and story characters.

- 1Large Language Models Meet Harry Potter: A Bilingual Dataset for Aligning Dialogue Agents with Characters腾讯AI实验室 · 2023年



