Research On Food Safety Information And Judgment System Based On Knowledge Graph Reasoning

Food safety is a matter of people's health and social stability,and how to conduct research and early warning of food safety events has always been a hot topic of social concern.With the development of the Internet,the amount of food safety information has exploded,and the forms and quality vary.Traditional intelligence research techniques have low efficiency and poor interpretability in the process of acquiring food safety information,and cannot provide good prediction and early warning for the occurrence of food safety risks.To address this problem,this paper uses dual process theory and knowledge graph inference techniques to build an efficient and interpretable food safety intelligence model,which can provide active decision support for early warning of food safety risks.This paper uses food safety sampling data from 2018-2020 and textual information from the book "Food Safety Accident Determination and Prevention and Control" as the basis of research,and the main work is carried out in the following aspects.(1)A food safety question-and-answer inference method based on multi-hop relations of knowledge graph is proposed.Firstly,we construct a food safety knowledge map as the basis of the question-answer inference model;secondly,we use graphical neural networks,path search and probabilistic attention mechanisms to construct a multi-hop inference question-answer model based on the constructed food safety knowledge map;finally,we establish a cross-entropy loss model between each pair of "question-answer" pairs and use dynamic programming Finally,a cross-entropy loss model is developed for each "question-answer" pair,and the answer entity in the corresponding path sequence is solved by dynamic programming.The constructed food-safe multi-hop reasoning model greatly improves the resolution and reasoning of complex questions in Q?A systems.(2)A food safety question-and-answer reasoning method based on cognitive mapping is proposed.Combined with the two-channel theory,the food safety cognitive reasoning model is constructed,and the cognitive way and solution ideas of human facing complex problems are applied to the training of the model.The model uses BERT(Bidirectional Encoder Representations from Transformers)pre-trained model and Graph Neural Networks(GNN)for implicit relationship extraction and explicit logical inference of food safety related information,and reasoning from both semantic and graph.The highest answer probability entity in the text document is inferred from both semantic and graph perspectives.This method further optimizes the QA system,improves the accuracy and interpretability of the QA,and provides a reasonable theoretical support for the diagnosis of food safety risk situation,and makes the QA system more credible.(3)Based on the two inference algorithms proposed above,a food safety question and answer system based on knowledge graph inference is built to provide decision support for food safety intelligence research and evaluation.