CocoChorales-E

该数据集是一个为训练和评估音乐错误检测模型而设计的合成音乐错误数据集。该数据集在类别上是不平衡的，为了解决这种不平衡，在训练过程中使用了加权的交叉熵损失。在规模上，平均错误与正确比例为1:4，所面临的任务是音乐错误检测。This is a synthetic musical error dataset designed for training and evaluating musical error detection models. The dataset is class-imbalanced. To mitigate this imbalance, weighted cross-entropy loss is utilized during the training process. In terms of scale, the average ratio of erroneous samples to correct ones is 1:4, and the core task of this dataset is musical error detection.