DuLeMon (Baidu Long-term Memory Conversation)

DuLeMon 是一个大规模的中文长期记忆对话数据集，它模拟长期记忆对话，专注于在长期交互中主动构建和利用用户和机器人角色的能力。 DuLeMon 包含大约 27.5k 人与人对话、449k 话语和 12k 角色基础句子。该语料库可用于探索长期记忆对话、个性化对话和角色提取/匹配/检索。

DuLeMon is a large-scale Chinese long-term memory dialogue dataset. It simulates long-term memory dialogues, focusing on the capability of actively constructing and leveraging the personas of both users and robots during long-term interactions. DuLeMon contains approximately 27.5k human-to-human dialogues, 449k utterances, and 12k persona base sentences. This corpus can be used to explore long-term memory dialogue, personalized dialogue, as well as persona extraction/matching/retrieval.

搜集汇总

背景与挑战

背景概述

DuLeMon是一个大规模中文长期记忆对话数据集，包含27.5k对话和449k话语，专注于长期交互中的用户和机器人角色能力构建，适用于长期记忆对话和个性化对话研究。由北京邮电大学、哥伦比亚大学和百度于2022年联合发布。