Design And Implementation Of Intelligent Training Platform

With the application of artificial intelligence technology,more and more fields begin to implement intelligence.For financial,insurance and other enterprises that need to invest a lot of human and material resources for staff training,intelligent training is more urgent.Therefore,combined with the research direction of the laboratory and the actual business needs,this paper designs and implements an intelligent training platform.Through the intelligent training of enterprise employees,it not only provides convenient and efficient training methods for employees,but greatly saves the investment in employee training.The research content of this topic mainly includes the design and implementation of the online and offline end of the intelligent training platform.The online processing of this platform is mainly through multi-dimensional calculation of the answers and standard answers of the trainee users,to get the percentile score and the total score of the trainee users' answers in each dimension.Among them,the online processing mainly includes five modules: the general preprocessing module converts the input text to half angle,cuts sentences,eliminates meaningless symbols and word segmentation to get the cleaned structured data;The accuracy calculation module calculates the word,word and sentence dimension similarity score and the total accuracy score of the text;The politeness analysis algorithm module calls neural network algorithm to calculate the politeness score of the text and check the feature words;In the logic analysis algorithm module,the sliding window is used to segment the text,and then the neural network model is used to calculate the logic score of the text segment list;The coherence analysis module first extracts the features of modal particles and single sentences,and then calculates the score;Finally,the total score is obtained by integrating the percentile results of each module.The off-line processing of this platform mainly focuses on the neural network models such as politeness analysis algorithm and logic analysis algorithm,such as data crawling,data cleaning,data annotation,model training,iterative optimization,etc.,which provides a high-performance and effective available model for the on-line processing of the platform.After the completion of the platform,the system test is carried out,and the functions of each module are in line with the expected effect.The average time for a single request of the total interface is 140 ms,and the accuracy rate of the test set is 93.45%,which meets the business requirements.At the same time,the platform is compiled,deployed and delivered by image.Now,the platform has been delivered to customers and put into use.