Verification of Relational Database Languages Codes Generated by ChatGPT

Abstract

The potential of using large language model artificial intelligence systems to generate program codes for application development is significant. Database codes in SQL (Structured Query Language), which is the standard relational database language, can be generated by such systems. Generative AI systems know database languages syntax through their training data and the text patterns from various sources that include SQL queries and related text. Thus, the generated codes may not be perfect and need verification before usage. This paper verifies the relational completeness of the SQL codes generated by ChatGPT, one of the most widely used large language model systems. Relational algebra operators are used for the relational complete verification. An equivalent relational calculus statement is generated for each SQL and relational algebra statement. The results confirmed that ChatGPT has the ability to generate relational complete SQL, relational algebra, and relational calculus codes.