基于粒计算的程序题评分模型研究

程序题自动评分是程序设计类课程实现无纸化考试和在线练习的核心问题之一，传统评分方法没有从程序的结构、语法、相似度等方面进行全面考虑。受粒计算相关理论及模型启发，结合人工评分的多粒度思维模式．对程序源码进行静态分析，提出一种基于粒计算思想的程序题自动评分模型。以C程序为例，将程序文本和标准答案程序在各个粒度层面上进行匹配，计算出各粒度权值，将匹配结果进行综合评价得出最终评分。实验表明，该评分方法全面考察了程序在结构、语法、文本相似度等多方面的匹配程度，有较高的准确性，并可方使地应用到其他类型程序的评分模型之中。

Research on Scoring Model of Program Based on Granular Computing

Automatic scoring for program is one of the most important issues of designing the online examination and practice system about the program designing courses. Traditional methods did not fully consider the program structure, syntax and similarity. Inspired by the theories and models of granular computing inspired, this paper proposes an automatically scoring model for program based on granular computing. The multi-granularity thinking mode of artificial scoring is adopted in this model,and the program codes are analyzed statically. C programs are used as examples. First, the program code and the standard answer code are matched at all levels. Then, the weights are calculated for each granularity. Last, the final score is obtained by matching the previous results. The experiments show that the proposed method has higher accuracy and a comprehensive consideration of the program structure, grammar, and text similarity. In addition, the method may be easily applied to scoring other programs.