Automatic Assessment of Programming Exercises Using Syntactic Analysis

This paper proposes a method to grade programming exam questions automatically. Our motivation is that there are no robust and scalable automatic methods for the analysis of computational thinking from source code programmed by elementary-level students. The approach to this problem supports the improvement of PC development in primary and secondary school students. The validation of the method is performed through the assessment of the answers of primary school students to programming exercises using a programming language called LIE++. The method assesses student answers using several techniques such as the analysis of programming structures, code clones and the execution of code based on input and output values defined during the specification of the exercises. The use of these techniques provides specific scores to obtain a grade of the student's answer. The source code analysis and scoring of exercise answers is carried out using high-performance computing for improving system response time. This research contributes a scalable method for the automatic evaluation of exams, which we expect to support the development of PC.