程序自动评分系统研究.pdfVIP

中国科技论文在线 Research on automatic programs grading system ** Yang Meng, Zhao Xinyan 5 (Beijing university of posts and telecommunications, Economic management institute, Beijing 100876) Abstract: This project is a research one, which focus on is to research on automatic marking method to realize automatically marking Java program. In the process, generated the story cards of the semi-auto marking system, gave them the implementation sequence, determined which story cards were more 10 important and which ones were not that important and could be realized after the important ones. This paper deeply researched on four marking methods, method based on Program Understanding; method based on the Artificial Intelligence; method based on Sentence Similarity; and method based on String Similarity. Evaluated them, knew each of their advantage and disadvantage and then got good ideas from them. Then generated own method, which used the method based on String Similarity and gave 15 some improvements to it. The method can successfully check if the program can be compiled automatically; check if the output is correct automatically; check if the source code is written according to the spec optionally, which means that it can choose to check automatically or manually, and the system can check the code quality manually. After finished the system, this paper did some evaluations to check whether this method is good. This paper used three different kinds of programs to evaluate 20