Implementation And Optimization Of The Questionnaire Automated Audit System

Science popularization survey is an important tool for government to master the scientific quality of the people,and it is also an important basis for the government to formulate related policies to improve scientific quality.The household questionnaire survey is the main way for science popularization agencies to collect various science popularization survey data.The standardization of the survey process will affect the reliability of the data obtained,and then the scientificity of related decisions.At present,the review of the standardization of the survey process mainly depends on the manual review of the questionnaire information.Manual review consumes a lot of human and material resources,and the review time of a single questionnaire is relatively long and inefficient,which cannot meet the increasing survey Review requirements of the questionnaire.The use of an automated questionnaire review system to assist relevant personnel in the review of the questionnaire can reduce the pressure on the reviewers and increase the efficiency of the review questionnaire.However,there are relatively few studies on the questionnaire automated review.In some existing studies,the questionnaire review also has many problems such as limited scope of application and failure to provide explanations for the causes of errors.To this end,this article conducts a detailed analysis of the possible problems in the questionnaire,designs and implements a set of automatic questionnaire audit system,reviews the eligibility of the questionnaire from voice,image and GPS,and uses The questionnaire data was tested.After testing,the average time to review a questionnaire is about 1 minute,the efficiency is much higher than the efficiency of manual review.Specifically,the main work of this article includes the following aspects:(1)Use voice technology to improve the accuracy of the questionnaire review.Through voice recognition technology,convert the recorded information in the investigation process into text information,and calculate the text similarity according to the text of the relevant topic,determine the wrong topic number and type according to the set threshold,and use silent detection,Chinese pinyin correction The method improves the efficiency and accuracy of the system.Using audio analysis,the AUC value of each sub-question in the questionnaire reached 0.95,and the Precision,Recall,and F1 values reached 0.96,0.93,and 0.94,respectively.(2)Use image recognition technology and GPS technology to improve the accuracy of the questionnaire review.For the randomly captured images in the questionnaire,the image module uses the face detection method to identify the number and gender of the images to determine whether the survey environment and the gender of the interviewee meet the requirements.The geographic location module uses the GPS function to analyze the geographic location of the survey location and determine whether the investigator went to the designated neighborhood committee to investigate as required.The data shows that the image module can review about 0.8% of the gender cheating problems that have not been discovered by manual review.The geographic location module has a recall rate of 100% of the geographic location cheating problems that have been manually reviewed and can find more potential cheating Questionnaire.(3)In the actual system,a set of automatic questionnaire audit system is implemented.The original background management system has been modified,and a set of automatic questionnaire audit system containing the above functions has been implemented and deployed on the Alibaba Cloud server.Related audit pages are provided to facilitate relevant personnel to query the cause of the error.The use of multi-threaded parallel processing,IP address detection and other methods ensure the effectiveness and reliability of the system.