Research And Implementation Of Intelligent Evaluation Technologies For SQL Exercises

With the development of Internet technology,Massive open online course(MOOCs)have become an important form of education and attracted increasingly more attentions.Database technology is one of the most important courses of computer science.Therefore,mainstream MOOC platforms at home and abroad provide online courses of database technology.Learning SQL is the key to master database technology.On the one hand,like other programming languages,learning SQL is based on programming practice,which means students’ are required to do a large number of SQL exercises.On the other hand,there may be thousands or even tens of thousands of students are enrolled in an online course.Hence,how to realize the intelligent evaluation of large-scale SQL exercises has become a hot research topic in recent years,which involves two aspects:partial marking of SQL exercises and detection of SQL plagiarism.However,there are still some shortcomings in existing research works,which are reflected in the following two aspects: 1)the partial marking technologies of SQL exercises do not consider the diversity of exercise answers,which reduces the fairness of evaluation;2)the detection technologies of SQL plagiarism do not make good use of the syntax and coding characteristics of SQL language,which reduces the accuracy of detection.To this end,this thesis studies and implements several intelligent evaluation technologies for SQL exercises,aiming to improve the fairness of partial marking for SQL exercises and the accuracy of plagiarism detection for SQL exercises.The research contents of this thesis include:(1)This thesis proposes a novel partial marking technology for SQL exercises based on the equivalent transformation of SQL answers(named SQL-GRADER).On the one hand,SQL-GRADER uses equivalent SQL answers obtained by the equivalent transformation of a SQL answer to ensure the fairness in marking students’ exercises with different forms of answers;on the other hand,it puts forward a correction strategy for students’ answers which have some errors and a method of quantifying students’ scores based on the correction cost and the conversion cost between the corrected answer and the correct answer.Then,SQLGRADER is evaluated through a real classroom practice participated by several teaching classes.Experimental results show that SQL-GRADER has more advantages than the state-of-the-art technology in terms of the marking fairness,and the marking precision of SQL-GRADER is on average improved by 33%,compared with state-of-the-art technologies.(2)This thesis proposes a plagiarism detection technology for SQL exercises based on coding features of students.Firstly,based on a student’s SQL code to an SQL exercise which will be judged,the student’s coding features w.r.t.the exercise(such as the database schema feature)and the generalization coding features w.r.t.the exercise(such as the SQL keyword writing habit in terms of upper/lower case)are extracted.Then,through the hierarchical clustering analysis of the coding features of all students w.r.t.the current exercise,the potential plagiarism groups are derived.For each student in a plagiarism group,the similarity between the student’s generalization coding features w.r.t.the exercise and historical generalized coding features extracted based on the studnent’s historical SQL codes is computed,which will then used to identify the copier and the giver: if the similarity value is low,the student deemed as a copier;otherwise the student is deemed as a giver.The proposed technology is verified by using the dataset collected in a real classroom practice in which students are asked to do a group of SQL exercises online.Experimental results demonstrate that the proposed technology successfully improves the accuracy of plagiarism detection for SQL exercises by 10.22%,compared with related technologies.The intelligent evaluation technology for SQL exercises is a problem worth further study,having important theoretical and practical significance.In this thesis,we propose a novel partial marking technology for SQL exercises based on equivalent transformation and a detection technology for SQL plagiarism based on coding features of students.Experiments based on datasets derived from real teaching practices verify the effectiveness of the two proposed technologies.