Assessment Of Learning Achievement In Intelligent Tutoring System

Intelligent Tutoring System is a computer learning environment,which integrates the contents of cognitive science,computer vision,natural language processing and other fields.The ultimate purpose is to improve students’ knowledge level and learning ability without human intervention.Specifically,the Intelligent Tutoring System can monitor the interaction between students and the system in real time,obtain students’ data and analyze the learning situation,automatically provide appropriate learning content to students according to students’ ability level and style,and complete the whole teaching process through evaluation feedback,interactive communication in class and intelligent homework recommendation after class.This series of functions of Intelligent Tutoring System are inseparable from the assessment of students’ learning achievements.A reasonable evaluation mechanism can reduce students’ repeated and meaningless learning,provide students with accurate and timely feedback,and improve the teaching quality to a greater extent.From the perspective of application,the current evaluation technology can be divided into two categories: the evaluation of students’ learning performance and in-depth cognitive diagnosis.Student performance evaluation is to give students an overall score,and cognitive diagnosis needs to further infer students’ strengths and weaknesses and their mastery.Students’ performance evaluation usually uses the theory of fuzzy sets,but it has some obvious disadvantages: it depends on the time of students’ answer to each question,can not distinguish the students with the same score,the method is complex and can not reflect the idea of instructional design.In terms of cognitive diagnosis,the traditional cognitive diagnosis methods simplify cognitive modeling,resulting in the diagnosis result is binary,master or not master.In addition,some advanced cognitive diagnosis models have complex parameter estimation methods and high time complexity,which is not conducive to the real-time interaction with students.In view of the shortcomings of performance evaluation,we design a new fuzzy reasoning method to determine students’ grades.Specifically,first define appropriate fuzzy sets fuzzify the three attributes of problem: importance,complexity and difficulty,calculate the degree that the three attributes belong to the three levels of low,medium and high,and then infer the comprehensive level of a problem according to the designed reasoning rules.A calibrated accuracy can be calculated for each student from the accuracy of the original answer,and the student’s score can be recalculated.Aiming at the shortcomings of cognitive diagnosis,this paper proposes a cognitive diagnosis model based on Bayesian network.Firstly,the item response theory is introduced to simplify and determine the conditional probability distribution,then the educational hypothesis is proposed to simulate students’ cognitive processing process,and finally the error coefficient and guess coefficient are introduced to make the model more accurately simulate the response of examinee to questions.We only need to estimate the prior probability distribution of tested group’s ability which the student belong to,and collect the test results to automatically update the Bayesian network to infer the distribution of student’ability and related skill.By predicting students’ learning performance on real data sets,the effectiveness of the diagnostic model is illustrated,and a specific application is given to show its good interpretability.