Chinese short text classification method based on prompt learning

The invention discloses a Chinese short text classification method based on prompt learning, which comprises the following steps of: 1) automatically searching a template in a continuous space by using a small number of parameters by using an unnatural language template, and reducing the cost and error of manually designing the template; 2) category prediction: sending the new input data added with the template to the selected pre-training language model, and performing category prediction on the input data by using the bert pre-training language model; and 3) after the probabilities of all the label words are calculated, mapping the predicted probability on each label word to a real label, and finally obtaining a classification result. According to the Chinese short text data set classification method, the accuracy of Chinese short text data set classification is improved, good classification can be achieved only through few training samples, dependence of the model on large-scale data is greatly reduced, th