Click bait detection method for prompting learning only by utilizing title

The invention discloses a click bait detection method for prompt learning only by utilizing a title in the field of natural language processing research, which comprises the following steps of: 1, selecting a proper pre-training language model as a trunk, and constructing a tag word and a template in prompt learning; 2, optimizing label words in prompt learning through five optimization strategies, and converting a classification task into a probability calculation problem of category label words by utilizing the expanded label words; 3, constructing an input text and a preset prompt template into a prompt text with mask, taking the prompt text as the input of the model, and carrying out click bait detection by utilizing the optimized tag word; 4, finally mapping the probability predicted on each label word to a corresponding category to obtain a final predicted score of the label as a classification result; according to the method, the label words are screened, prompted and learned by using five optimization