Bridge text small sample named entity recognition method based on prompt learning

The invention discloses a bridge text small sample named entity recognition method based on prompt learning, and the method specifically comprises the steps: analyzing to-be-tested text data of a bridge engineering database, and carrying out a small amount of manual marking; corresponding prefix prompt words and label prompt words are constructed for the marked corpora and the unmarked corpora through input statement templates and label sequence pairs; combining the two types of prompt words and inputting the combined prompt words into a pre-training language model for coding to obtain comprehensive representation of the prompt words; and training a corresponding model by calculating a total loss function, and outputting an entity tag corresponding to the lexical element in the to-be-tested statement. The problems that in a named entity recognition task in the bridge engineering field, a large number of labeled training samples are needed, the model training efficiency is low, and the final entity classificat