Design Of Intelligent Tutoring System Based On Deep Learning

The online education of artificial intelligence is one of the emphases of education system research and development recently.Compared with traditional teaching methods,online education provides students with a freer learning environment and richer teaching resources.Companies such as xueersi,squirrel AI,new Oriental and ape test bank are all making great efforts to develop online education.At present,there are two bottlenecks restricting the development of intelligent online education system :(1)a large number of offline students' knowledge,ability and behavior feature materials are not informationized and cannot be used for online education;(2)the online practice system lacks self-adaptation,and it cannot recommend personalized practice contents by comprehensively considering students' knowledge,ability,behavior characteristics and current practice results.Aiming at the above two problems,this paper adopts deep learning technology to implement a computer vision recognition system and an adaptive practice recommendation system for offline student evaluation materials.Specific contributions are as follows:(1)The deep learning computer vision technology,design and implement a set of the results of the offline student assessment materials and answer the questions on the visual recognition system,solved the student assessment identified within the student handbook and students answer the problem of handwritten digit recognition,student handbook recognition accuracy of 95% or more,the students of handwritten numerals recognition accuracy of 80%.The system is already in commercial use at a large Internet education company.(2)Based on the MIT open source online practice system CAT-SOOP,a set of adaptive practice recommendation system based on deep reinforcement learning is designed and implemented,which can provide students with adaptive practice recommendation based on student practice data,training student knowledge tracking model and reinforcement learning recommendation engine.The system has completed the prototype design,implementation and preliminary verification.The offline student information depth visual recognition system we designed is universal and can be widely used in other similar scoring application scenarios.The self-adaptive practice recommendation system designed by us has certain exploration and innovation,which provides beneficial reference for the progress of the whole industry,and has important practical significance for saving students' learning time,improving students' learning ability and improving the teaching quality of online education.29 Figures,5 tables,39 references.