ShareChatX
收藏资源简介:
ShareChatX是由浙江大学和美团联合创建的大规模语音对话数据集,旨在解决现有语音对话数据集在规模和场景多样性上的不足。该数据集包含947,236条对话,涵盖了情感、音频事件和音乐等多种复杂场景。数据通过大语言模型生成对话脚本,并结合可控的文本到语音合成模型生成语音对话。数据集的应用领域主要集中在提升语音对话系统在复杂场景中的表现,特别是在涉及音频和音乐的场景中。通过合成数据与真实数据的结合,ShareChatX为训练更强大的语音对话模型提供了丰富的数据支持。
ShareChatX is a large-scale spoken dialogue dataset jointly created by Zhejiang University and Meituan, aiming to address the shortcomings of existing spoken dialogue datasets in terms of scale and scenario diversity. This dataset contains 947,236 dialogue sessions, covering various complex scenarios such as emotions, audio events and music. The dialogue scripts are generated by large language models and combined with controllable text-to-speech (TTS) synthesis models to produce the final spoken dialogues. The primary application of this dataset is to enhance the performance of spoken dialogue systems in complex scenarios, especially those involving audio and music. By combining synthetic and real-world data, ShareChatX provides rich data support for training more robust spoken dialogue models.

- 1OmniChat: Enhancing Spoken Dialogue Systems with Scalable Synthetic Data for Diverse Scenarios浙江大学, 美团 · 2025年



