WikiMuTe

WikiMuTe是由华为技术有限公司慕尼黑研究中心和Universitat Pompeu Fabra的音乐技术组合作创建的一个公开可用网络源数据集，专注于音乐音频的语义描述。该数据集从维基百科的丰富文章目录中提取，涵盖了音乐作品的广泛信息，如流派、风格、情绪、乐器和速度等。通过专门的文本挖掘管道，数据集收集了长形式和短形式的描述，旨在通过深度学习模型实现文本与音频的匹配，支持跨模态检索和音乐自动标记等应用。

WikiMuTe is a publicly available web-sourced dataset developed through a collaboration between the Munich Research Center of Huawei Technologies Co., Ltd. and the Music Technology Group at Universitat Pompeu Fabra, focusing on semantic descriptions of musical audio. This dataset is extracted from the extensive article catalog of Wikipedia, covering a wide range of information about musical works, such as genre, style, emotion, instruments, tempo, and more. Leveraging a dedicated text mining pipeline, the dataset curates both long-form and short-form textual descriptions, aiming to enable text-audio matching using deep learning models, thereby supporting applications including cross-modal retrieval and automatic music tagging.