Aims to design a question-answering system in the military domain based on the basic pre-trained models of spa-Cy.Firstly,a method for entity recognition is introduced,which extends and fine-tunes the model using similarity assessment.Sub-sequently,the integration of the Matcher component into the spaCy pipeline,coupled with the Adam neural network optimizer,is explored for extracting entity relationships.Validation shows that the aforementioned approach improves the recognition accuracy of entities and entity relationships compared to the basic pre-trained model of spaCy,addressing the issue of insufficient labeled data in the military domain effectively.This provides a reliable reference for the development and application of question-answering sys-tems in the military domain.
维普
万方数据