Research On Network Health Information Credibility Assessment Model Based On Multidimensional Information Characteristics

Health has always been the subject of people’s attention,and people’s demand for health information has always been very strong.In the past,public access to health information was mainly through consultation with hospital doctors and reading health magazines or newspapers.With the popularity of mobile Internet,the way the public gets health information is also changing.Medical workers,health institutions,ordinary patients,medical enterprises and government agencies are all online health information providers.Abundant online health information makes it convenient for the public to retrieve disease information,but it is also full of false information,exaggerated information,one-sided information and so on,which is easy to mislead the general public who lacks medical and professional knowledge,and may harm health in serious cases.It is difficult to judge the accuracy of content by human,but to judge a large amount of information.Therefore,it is of great significance for the general public and the network health community to realize the method of automatically judging the credibility of network health information.In this study,an identification model of network health information reliability is constructed,and the reliability of network health information can be measured automatically and quantitatively based on information characteristics.In this study,an index system for credibility recognition was designed based on the theory of information transmission from three aspects: information content characteristics,information source characteristics and information transmission characteristics.Secondly,the quantitative measurement method of each indicator is determined.Natural language processing technology is used to measure information content text and assignment transformation is used to measure other information features.In order to verify the effect of the model,health information samples of different content categories were selected from the network to verify the Kendall consistency between the model results and the results marked by experts,as well as the performance of the classifier.It is verified that compared with expert annotation,the reliability evaluation results of this model are consistent with above 0.9,0.7 and 0.8 respectively,and the credibility classification of coarse-grained and the credibility ranking of fine-grained have similar consistency,indicating that the evaluation model is close to the expert level and applicable to differentgranularity.As a classifier,the accuracy of information reliability is up to 0.944,and the ability to distinguish untrusted information is strong.This study effectively evaluates information credibility from the information feature dimension and makes use of publicly available data to conduct automated evaluation,which is more efficient and less costly than expert manual evaluation and can assist information screening in platform supervision and public utilization.Research on mining and service of big data resources--for medical and health fields,the MOE Project of Key Research Institute of Humanities and Social Sciences at Universities(No: 17JJD870002)...