【摘要】 目前对编程题的自动评阅技术多建立在程序能正常运行的基础上。实践表明,这样的评价基础并不适用初学计算机程序语言的学生,大量的答题由于很小的编译错误而被判为零分。如何对一个无法正常运行的程序文本给出一个正确度的评价,这个课题的研究在国内外还未见有特别有效的方法。把文本相似度和程序相似度的研究成果应用于对程序文本正确度的评价上是一种新的探索。通过构造由样本程序生成的目标程序集,结合使用衡量文本相似度的Levenshtein距离,给出一个程序正确度评价方法。希望在程序文本的正确性评价方法的研究方面能有所突破。更多还原

【Abstract】 Currently automatic scoring technology for programming tests is based mainly on whether the programme runs properly.Experience tells that such scoring technology doesn’t fit for beginner programmers as a great deal of answers are scored 0(zero) for some subtle errors.It still remains a puzzle how to make a proper assessment to a segment of faulty codes.There is no effective way to solve the problem both at home and abroad.Applying research results of text similarity and programme similarity to code correctness assessment is a novel exploration.Through generating a target programme aggregation from sample programmes and utilising Levenshtein distance for text similarity measurement,a programme correctness assessment approach is put forward which is deemed to strive for a breakthrough in the research on code correctness assessment.更多还原