Abstract: Based on the teacher- student model, a semi-automatic annotation method for datasets was proposed, which solved the problem of large workload of dataset manual annotation, different data quality and high professional threshold in supervised learning. In the cloud experiment, the annotation method was used to realize the semi-automatic labeling of the clock synchronization pattern classification data. On the other hand, the automatic evaluation of the difficulty of the data set was realized, which can be used to guide the optimization and evaluation of the machine learning model.

Key words: machine learning, data annotation, teacher-student model