Research And Implementation Of Public Opinion Early Warning Model Based On Semantic Relation Diagram

In recent years,frequent online public opinions have brought new challenges to the timeliness and accuracy of public opinion early warning.First of all,it is a challenge based on multi-source data.Multi-feature integration factors such as standardization,colloquialism and mixed text of short and long text brought by multi-source mixed text,such as news websites,network communities and social media,lead to poor universality of public opinion model and low early-warning accuracy.Secondly,public opinion warning model based on keywords or shallow semantic calculation leads to low warning accuracy.Real-time monitoring of multi-source complex text,and improve the accuracy of public opinion early warning,to provide decision data support for the security governance of digital content in cyberspace.Multi-source mixed text data characteristic are analyzed in this paper,combined with field in university,design contains the attributes of the public opinion field semantic nodes and semantic relation,construct the semantic relation graph model,combines the technique of map database related to realize the semantic relationship diagram in the field of automated build,implement semantic node automatic expansion model and semantic relations.Based on semantic field diagram,design contains lexical semantic computing,lexical semantic relevance semantic computing,text,text normalization calculation,computation of early warning level of early warning model,multidimensional,deep effectively the hybrid text of the source data are multi-level public opinion semantic analysis and grading warning.Based on the public opinion early warning model of semantic relation graph,this paper designs and implements a multi-source data public opinion early warning system,which provides functions such as data source collection,data persistent storage and front-end visualization of early warning results,etc.,which can effectively help the government and relevant public opinion departments to timely monitor and master network public opinion.In order to verify the effect of public opinion warning model,multiple real data source network texts are taken as experimental data,and the research results of similar algorithms are selected as comparative experiments.The experimental results show that the accuracy F value of the public opinion early warning model in this paper can reach 85.14%,which is more than 7.8% higher than the reference model,and can meet the needs of the real network public opinion early warning and monitoring.