Research And Implementation Of Intelligent Analysis Method For Food Safety Events

In recent years,China has implemented a series of governance policies for food safety,and the food safety situation tends to be stable.But with the advent of the era of big data,once appear food safety events will be spread on the Internet,such as happened in recent years "salmon found on board will be coronavirus","a food and beverage company in tianjin catering environment considered the" and "sauerkraut in graves preserved" food safety incidents,sparked intense discussion of Internet users in the network.The explosive growth and dissemination of information make it more difficult to clarify the ins and outs of events,affecting the supervision and management of food safety events.Therefore,it can be seen that clarifying the development process of food safety incidents,fully mining event information,and comprehensively analyzing food safety incidents play an important role in food safety incident supervision,government decision-making,and solving people’s livelihood issues.In the early stage,event analysis was only carried out from the perspectives of causality,timing and event chain among events,and the event information obtained was not comprehensive.Subsequently,the proposal and application of knowledge graph,especially the emergence of event-centered matter graph,provide support for the intelligent analysis of events and the study of the evolution law of events.However,there is still room for improvement.For example,the ability to extract syntactic structure information during event extraction is insufficient,and the incomplete mining of sentence hierarchy structure information during event relation extraction leads to the inaccuracy of the constructed event atlas.To solve the above problems,this paper uses graph convolutional neural network to extract events,ON-LSTM and attention mechanism to extract event relationships,and builds a food safety event mechanism map on this basis,and designs and implements a food safety event analysis system.The main work and innovations of this paper are as follows:(1)This paper collected food safety event texts in recent years and defined a set of food safety event text labeling rules to label the collected food safety event texts and build a food safety event text corpus.(2)An event extraction method based on graph convolutional neural network and word fusion is proposed.In this method,word feature vectors containing syntactic structure information are extracted by graph convolutional neural network with dependency,and the word feature vectors are fused with word feature vectors to obtain richer sequence representation information.(3)An event relation extraction method based on ON-LSTM and attention mechanism is adopted.This method uses ON-LSTM,which can better capture the structural level information in sentences,and further strengthen the important information through the attention mechanism,so as to better extract the relationship between events.The event relation pairs obtained from the event extraction and event relation extraction tasks were stored in the Neo4 j graph database to construct the food safety incident mechanism atlas.On this basis,a food safety incident analysis system is designed and implemented,which has the functions of keyword query,information extraction,event heat analysis and custom atlas.