AliCHI
收藏资源简介:
AliCHI是由阿里巴巴集团达摩院创建的大规模多模态数据集,专注于人类对人类面对面交流的精细标注。该数据集包含635个对话会话,总时长52小时,来源于200名参与者。创建过程中,通过设计通用话题和确保参与者具备基本知识,以减少个体偏差。数据集主要用于支持人机交互系统中的背通道预测和轮流预测任务,旨在提升对话系统的自然性和人类化行为。
AliCHI is a large-scale multimodal dataset developed by Alibaba Group's DAMO Academy, focusing on finely annotated human-to-human face-to-face conversations. The dataset includes 635 dialogue sessions with a total duration of 52 hours, sourced from 200 participants. During its construction, general conversation topics were designed and participants were required to possess relevant basic knowledge to mitigate individual biases. This dataset is primarily used to support backchannel prediction and turn-taking prediction tasks in human-computer interaction systems, aiming to improve the naturalness and human-like behaviors of dialogue systems.

- 1AliCHI: A Large-scale Multi-modal Dataset and Automated Evaluation Tool for Human-like Dialogue Systems达摩院 · 2022年



