Small sample sentence pattern classification method, system and equipment based on task prompt and medium

The invention discloses a small sample sentence pattern classification method, system and device based on task prompt and a medium, and the method comprises the steps: classifying a data set, collecting, cleaning and marking the classified data set, and forming a sentence pattern data set; on the basis of an existing model, modifying the model to obtain a modified model; training the modified model by using the sentence pattern data set to obtain a trained model; and carrying out effect verification on the trained model, taking the model with the best effect as a final model, and classifying the input small sample sentence patterns by utilizing the final model. The new deep learning method is used, the improved PET model BEK-PET is used for evaluating the method in Chinese sentence pattern classification, the optimal method is found out, and the robustness of the classification model is improved. 本发明公开了一种基于任务提示的小样本句型分类方法、系统、设备及介质，其方法包括：对数据集进行分类，对分类后的数据集进行收集、清洗和标注，形成句型数据集；在现有模型的基础上，对模型进行修改，得到修改后的模型；利用句型数据集对修改后