Question and answer method, system and method based on multi-hop knowledge graph and medium

The invention discloses a question and answer method, system and method based on a multi-hop knowledge graph and a medium, and the method comprises the steps: 1) inputting a question to be answered into a trained Sensor-BERT model, and obtaining a plurality of sentence initial vectors with fixed dimensions; 2) labeling the sentence representation vector, and inputting the labeled sentence representation vector into an FNN-based question representation learning model to obtain a question vector; and 3) obtaining a knowledge graph vector representation, and according to the knowledge graph vector representation and the question vector obtained in the step 2), performing answer selection in all answer candidate sets through a scoring function to obtain a multi-hop knowledge graph question-answer result. The method, the system, the method and the medium can perform a relatively accurate entity extraction task in the field of an inspection robot system; and outputting the selected best entity as an answer. 本发明公开了一