Design and Implementation of an Automatic Grading System for Programming Code Based on Artificial Intelligence

This paper proposes an automatic grading system for programming code based on artificial intelligence, aiming to improve the efficiency and accuracy of grading programming assignments. The system "Semantic-Structure Fusion Scoring Algorithm" (SSFS) combines code semantic understanding and...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:2025 IEEE 3rd International Conference on Image Processing and Computer Applications (ICIPCA) s. 1846 - 1851
Hlavní autor: Jiang, Guoyu
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 28.06.2025
Témata:
Accuracy
Artificial intelligence
automatic scoring system
Codes
Computer languages
Education
programming code
Programming profession
semantic-structural fusion scoring algorithm
Semantics
Standardization
Syntactics
Vectors
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:This paper proposes an automatic grading system for programming code based on artificial intelligence, aiming to improve the efficiency and accuracy of grading programming assignments. The system "Semantic-Structure Fusion Scoring Algorithm" (SSFS) combines code semantic understanding and structural analysis to accurately evaluate code quality. In terms of semantic understanding, natural language processing technology is used to convert the code into semantic vectors to capture the logical meaning of the code; structural analysis uses abstract syntax trees to analyze the integrity and standardization of the code structure. The experiment selected 1,000 programming assignments of different difficulty levels, covering a variety of programming languages. The results show that the average similarity between the SSFS algorithm score and the professional teacher score reached 93%, which is 25% higher than the traditional scoring algorithm based on keyword matching. The system scoring takes an average of only 3 seconds, which is much more efficient than manual scoring. At the same time, for assignments of different difficulty levels, the scoring accuracy is maintained above 90%, which fully verifies the effectiveness of the system and algorithm, and provides an efficient and reliable solution for programming teaching and evaluation.
DOI:10.1109/ICIPCA65645.2025.11139057