Publicación: EXPLORING COPILOT GITHUB TO AUTOMATICALLY SOLVE PROGRAMMING PROBLEMS IN COMPUTER SCIENCE COURSES
Fecha
2023
Título de la revista
ISSN de la revista
Título del volumen
Editor
IEEE CONFERENCIAS
Resumen
IN RECENT TIMES, THE FIELD OF COMPUTER PROGRAMMING HAS EXPERIENCED A SIGNIFICANT REVOLUTION, THANKS TO ADVANCEMENTS IN MACHINE LEARNING. APPLICATIONS HAVE EMERGED WITH THE CAPABILITY TO GENERATE SOURCE CODE FROM NATURAL LANGUAGE DESCRIPTIONS. THESE TOOLS PRIMARILY UTILIZE LANGUAGE MODELS BASED ON DEEP LEARNING, WHICH HAVE BEEN TRAINED ON A COLLECTION OF PROGRAMS AND PROJECTS HOSTED IN PUBLIC REPOSITORIES. ONE OF THESE TOOLS IS GITHUB COPILOT, AN ARTIFICIAL INTELLIGENCE CAPABLE OF GENERATING SOURCE CODE THAT CAN BE INTEGRATED AS AN EXTENSION INTO DEVELOPMENT ENVIRONMENTS. THE OBJECTIVE OF THIS STUDY IS TO EXPERIMENTALLY EXPLORE, ANALYZE, AND EVALUATE THE SUGGESTIONS MADE BY THE GITHUB COPILOT TOOL IN PROGRAMMING TOPICS RELATED TO THE COMPUTER SCIENCE DEGREE AT THE UNIVERSITY OF BIO-BIO. WE PROPOSE FIVE STEPS: (1) COLLECTING NATURAL LANGUAGE STATEMENTS FOR BOTH GENERAL AND SPECIFIC PROGRAMMING PROBLEMS; (2) UTILIZING GITHUB COPILOT TO GENERATE PROGRAMS; (3) EVALUATING ITS PERFORMANCE; (4) CONDUCTING AN ANALYSIS; AND (5) MEASURING CODE QUALITY. THIS APPROACH ALLOWS US TO GAIN AN INITIAL UNDERSTANDING OF ITS EFFECTIVENESS, EMPHASIZING ITS APPLICATION FOR WELL-ESTABLISHED PROBLEMS AND MONITORING ITS USE FOR PROBLEMS WITH DISTINCT OBJECTIVES.
