In recent years, prompt-based learning has achieved some success in various natural language tasks. In text classification tasks, the construction of prompt templates and label mapping have a significant impact on the results. Therefore, many methods for constructing label mappings have emerged, such as manual and automatic construction. However, manual label construction is time-consuming, expensive, and has low scalability, while automatic label construction requires a large number of samples and other limitations. Therefore, we have explored a automatic label mapping method that only requires one sample. On the basis of only requiring one sample, we achieved good experimental results. In this paper, we propose an effective one-shot method for automatically constructing label mapping, ALaM. Our method is based on statistical algorithms and label semantic information for one to many label mapping. Our experiments demonstrate that ALaM achieves competitive performance on the GLUE benchmark without human effort or external resources.