Zero sample event extraction system and method based on comparative learning and data enhancement

The invention discloses a zero sample event extraction system and method based on comparative learning and data enhancement in the technical field of natural language processing, and the method comprises the steps: obtaining a data set, and dividing the data set into known events and unknown events; rewriting the event description text of the unknown event; forming a positive and negative sample pair by the known event, the unknown event and the unknown event subjected to event description rewriting; inputting the positive and negative sample pairs into an event encoder to generate corresponding feature vectors; after a comparison loss function value is calculated based on the feature vector, model parameters in the event encoder are updated through gradient return; and performing classification and clustering based on the updated feature vector input output by the event encoder. According to the method, by comparing similar and heterogeneous samples, the annotation data of the known event and the unannotated