Shapter-level event argument extraction method and system based on difficulty sequence reasoning

The invention discloses a chapter-level event argument extraction method and system based on difficulty sequence reasoning, and belongs to the field of text information extraction. According to the method, a corresponding prompt learning template is selected according to the context of the document, the context and the prompt learning template are coded, a context representation and a prompt representation are obtained, and the prompt representation comprises a vector representation of an argument role; calculating a prediction difficulty score of each argument role according to the vector representation of each argument role, and sorting the argument roles according to the prediction difficulty scores to obtain a predicted reasoning path; performing information reasoning according to the sequence of the predicted reasoning paths to obtain reasoning probability distribution of each argument role; and according to the obtained reasoning probability distribution of each argument role, predicting the position of