|
Download README.md from rice404/SemanticVAD-Dataset: direct link, hf CLI and curl.
- Browser
- Download file 5.79 kB
-
https://huggingface.co/datasets/rice404/SemanticVAD-Dataset/resolve/main/README.md
- Command line
-
hf download hf://datasets/rice404/SemanticVAD-Dataset/README.md
-
curl -L -o README.md https://huggingface.co/datasets/rice404/SemanticVAD-Dataset/resolve/main/README.md
5.79 kB
| license: apache-2.0 | |
| size_categories: | |
| - 10K<n<100K | |
| language: | |
| - zh | |
| - en | |
| tags: | |
| - chat | |
| - dialog | |
| - full-duplex | |
| # SemanticVAD 对话状态检测数据集 🌟 | |
| ## 数据集简介 | |
| 本数据集为全双工人机语音交互系统提供语义级语音活动检测(Semantic Voice Activity Detection)任务的训练与测试支持,包含15,000条训练样本和4,400条测试样本,标注质量经过大模型验证优化。 | |
| ## SemanticVAD 💡 | |
| SemanticVAD 通过语义理解实现智能对话状态检测,通常由轻量级语言模型实现。 | |
| - 输入:人机交互文本(含历史与实时对话内容) + 当前发言人标识(`'human'`(用户)/`'agent'`(模型)) | |
| - 输出:四类控制标签 | |
| - 🗣️ human 发言时: | |
| - <完成>: 用户语义完全,模型可以开始回复。 | |
| - <未完>: 用户语义未完,模型继续等待用户输入。 | |
| - 🤖 agent 发言时: | |
| - <打断>: 用户试图抢夺话题主导权,模型需停止当前回复并聆听用户的新发言。 | |
| - <附和>: 用户赞同模型发言,模型可以继续输出。 | |
| ## 数据集结构 🗂️ | |
| ### 训练集(15,000条) | |
| - 数据分布及来源: | |
| | 标签类型 | 中文数据量(来源) | 英文数据量(来源) | | |
| |----------|------------------------------------------|----------------------------------| | |
| | <打断> | 3,000(MagicData-RAMC真实场景 + Ke-SpeechChat 多轮对话) | 1,500(Ke-SpeechChat 单轮合成对话拼接) | | |
| | <附和> | 3,000(MagicData-RAMC真实场景) | 1,500(CANDOR真实对话) | | |
| | <未完> | 2,000(Ke-SpeechChat 多轮合成对话) | 1,000(Ke-SpeechChat 单轮合成对话) | | |
| | <完成> | 2,000(Ke-SpeechChat 多轮合成对话) | 1,000(Ke-SpeechChat 单轮合成对话) | | |
| - 数据格式为alpaca,样例如: | |
| ```json | |
| { | |
| "instruction": "# Dialog\nHuman[历史]:怎么把人工智能技术用在虚拟现实开发上呢?\nAgent[实时]:将人工智能技术应用到虚拟现实开发中,可以通过智能算法来提升用户体验,比如使用机器学习来创建更真实的虚拟角色\nHuman[实时]:那具体是怎么实现的?比如说,如\n", | |
| "input": "", | |
| "output": "<打断>", | |
| "system": "# Role\n你是人机实时交互的**用户行为分析**模块,你将收到包含部分历史信息的 Human 和 Agent 最新实时对话记录 (Dialog)\n\n# 任务\n当前【Agent正在发言】,在此过程中,你需要基于对话分析 Human 的意图属于 <打断> 还是 <附和>\n\n# 输出\n不要有多余的分析,仅严格输出以下二者之一: <打断> 或 <附和>\n\n# 判断标准\n## <打断> 的情况\nHuman 行为: 试图抢夺话题主导权\n特征包括:\n- 提供新概念/词汇/判断(如命名、定性、对比)\n- 提出问题或异议\n- 引入与当前话题无关的新话题\n\n## <附和> 的情况\nHuman 行为: 赞同 Agent, 期望 Agent 继续说\n特征包括:\n- 使用零内容反馈(嗯/啊/对)\n- 机械重复 Agent 中的原词/同义词\n- 表达简单的确认或同意(如“是的”、“没错”)\n", | |
| "source": "Ke-多轮", | |
| "lang": "中文" | |
| }, | |
| ``` | |
| ### 测试集(4,400条) | |
| - 来源:Tencent Full-Duplex Spoken Dialogue Systems 测试集 | |
| - 数据分布:每类各1100条 | |
| - 标注保障:经Qwen2.5-72B-Instruct进行标签校验与修正 | |
| ## 原始数据来源 | |
| 本数据集融合以下公开资源,均已进行合规化处理: | |
| 1. MagicData-RAMC 数据集 | |
| ```Bibtex | |
| @article{yang2022open, | |
| title={Open Source MagicData-RAMC: A Rich Annotated Mandarin Conversational (RAMC) Speech Dataset}, | |
| author={Yang, Zehui and Chen, Yifan and Luo, Lei and Yang, Runyan and Ye, Lingxuan and Cheng, Gaofeng and Xu, Ji and Jin, Yaohui and Zhang, Qingqing and Zhang, Pengyuan and others}, | |
| journal={arXiv preprint arXiv:2203.16844}, | |
| year={2022} | |
| } | |
| ``` | |
| 2. CANDOR 数据集 | |
| ```Bibtex | |
| @article{reece2023candor, | |
| title={The CANDOR corpus: Insights from a large multimodal dataset of naturalistic conversation}, | |
| author={Reece, Andrew and Cooney, Gus and Bull, Peter and Chung, Christine and Dawson, Bryn and Fitzpatrick, Casey and Glazer, Tamara and Knox, Dean and Liebscher, Alex and Marin, Sebastian}, | |
| journal={Science Advances}, | |
| volume={9}, | |
| number={13}, | |
| pages={eadf3197}, | |
| year={2023}, | |
| publisher={American Association for the Advancement of Science} | |
| } | |
| ``` | |
| 3. Tencent Full-Duplex Spoken Dialogue Systems 测试集 | |
| ```Bibtex | |
| @misc{zhang2025llmenhanceddialoguemanagementfullduplex, | |
| title={LLM-Enhanced Dialogue Management for Full-Duplex Spoken Dialogue Systems}, | |
| author={Hao Zhang and Weiwei Li and Rilin Chen and Vinay Kothapally and Meng Yu and Dong Yu}, | |
| year={2025}, | |
| eprint={2502.14145}, | |
| archivePrefix={arXiv}, | |
| primaryClass={cs.CL}, | |
| url={https://arxiv.org/abs/2502.14145}, | |
| } | |
| ``` | |
| 4. Ke-SpeechChat 数据集 | |
| ```Bibtext | |
| @misc{KE-SemanticVAD, | |
| title={Advancing Speech Language Models by Scaling Supervised Fine-Tuning with Over 60,000 Hours of Synthetic Speech Dialogue Data}, | |
| author={Shuaijiang Zhao and Tingwei Guo and Bajian Xiang and Tongtang Wan and Qiang Niu and Wei Zou and Xiangang Li}, | |
| year={2024}, | |
| eprint={2412.01078}, | |
| archivePrefix={arXiv}, | |
| primaryClass={cs.CL}, | |
| url={https://arxiv.org/abs/2412.01078}, | |
| } | |
| ``` | |
| # Citation | |
| Please cite our Hugging-Face when using our code, data or model. | |
| ```Bibtext | |
| @misc{KE-SemanticVAD, | |
| author = {KE-TEAM}, | |
| title = {KE-SemanticVAD}, | |
| year = {2025}, | |
| publisher = {Hugging Face}, | |
| url = {https://huggingface.co/KE-Team/KE-SemanticVAD} | |
| } | |
| ``` |