关键词: 相似性分析, 自动评估, 分词, 树编辑距离, 正态分布

Abstract: In view of the difficulty and the diviation caused by the features of multifactor and fuzzy boundaries of the automated programming assessment model for SQL languages(SQLAPAM),this paper introduces an assessment model based on structure similarity matching.The overall framework of the model is proposed with the combination of static and dynamic assessment methods. After being processed by standardization and tokenization, the submitted SQL statements are transformed into the equivalent sequence of token pairs with which the model constructs a corresponding structure tree(Stree). Next the model calculates similarity between the acquired tree and the target tree using the tree edit distance improved in the cost model and the substructure contribution factor,and gains a similarity threshold. Finally, the model maps similarity to the score intervals with reference to the normal distribution theory and adjusts the deviation brought by the impact factors with the help of the similarity threshold. Meanwhile the final assessment result for the SQL program is provided.

Key words: similarity analysis;automated assessment;tokenization;tree edit distance;normal distribution