A Model to Automatically Evaluate Beginners’ Programs Based on Syntactic Knowledge Point and Deep Learning Technology
A Model to Automatically Evaluate Beginners’ Programs Based on Syntactic Knowledge Point and Deep Learning Technology
复制标题
DOI：
10.1109/scisisis55246.2022.10001960
复制
发表时间：
2022-11
期刊：
影响因子：
--
通讯作者：
Yumo Yan;Hirokuni Kurokawa
中科院分区：
文献类型：
--
作者：
Yumo Yan;Hirokuni Kurokawa
该报告提供了一个CNN模型，可以根据他们对计划概念的理解来自动评估初学者的计划。矩阵，由句法知识点（SKP）技术组成，用于培训数据，是根据开始的42个程序概念中的每个程序，通过手动评估数据库中的每个程序。例如，有监督的学习，数据标准化，数据增强也已被用于训练模型。
This report provided a CNN model to automatically evaluate beginners’ programs based on their understanding of program concepts. The model input is a beginner’s program and the output is a letter grade for each program. To train the model, each input program was encoded as a matrix, composed of positive integers by Syntactic Knowledge Point (SKP) technology. Labels for the training data were made by manually evaluating each program in the database, based on the beginner’s understanding of each of the 42 program concepts. In addition, several other technologies such as supervised learning, data standardization, data augmentation have also been used to train the model. Finally, a model with an accuracy of 0.93 has been produced, setting a new benchmark in this young field.