CareCall (CareCall for Seniors)

carecall 是一个用于满足角色的对话系统的韩语对话数据集。该数据集由几个人类书面对话样本组成，这些样本使用大规模 LM 的上下文少镜头学习。大型 LM 可以生成具有特定个性的对话，提示包含对聊天机器人属性的简要描述和少量对话示例。我们使用这种方法来构建整个数据集。

CareCall is a Korean dialogue dataset tailored for role-compliant conversational systems. It comprises multiple human-written dialogue samples, which are generated via in-context few-shot learning with large-scale language models. Large Language Models (LLMs) can produce dialogues with predefined personalities, where the input prompts include concise descriptions of the chatbot's attributes and a small set of dialogue exemplars. We adopted this approach to construct the entire dataset.

搜集汇总

背景与挑战

背景概述

CareCall是一个韩语对话数据集，用于角色特定对话系统，通过大规模语言模型的少镜头学习技术生成个性化对话。数据集由NAVER AI Lab发布，适用于小样本学习任务，采用CC BY-NC-SA 4.0许可。