Inteligência artificial na documentação de sistemas legados: uma abordagem comparativa de ferramentas

Abstract

Software documentation is an essential activity in software engineering, playing a crucial role in communication, maintenance, scalability, and user experience, and serving as a fundamental pillar for the success of any software development project. However, this documentation activity is often underestimated. Legacy systems play a critical role in the operations of many companies but frequently face significant challenges related to maintenance and documentation. This results in substantial financial costs for organizations and demands considerable time from programmers working with these systems. Automation powered by artificial intelligence offers an alternative to address these challenges by optimizing the documentation of legacy systems, reducing costs, and improving maintenance efficiency. Identifying the most effective AI tool for this purpose is essential for the adoption of more precise and efficient strategies in professional contexts, promoting improvements in system maintenance and extending their lifespan. In this study, a comparative analysis was conducted on code documentation generation tools, focusing on their application to legacy systems. The comparison was based on the insertion of selected source code excerpts into the tools, with the objective of evaluating the generated documentation according to criteria such as readability, length, and adequacy of comments. The results indicated that the Gemini tool demonstrated the best performance in the documentation generation task; however, other tools ChatGPT, DeepSeek, Claude, Meta AI, Le Chat Mistral, and Grok also provided relevant results, highlighting the potential for improvements both in the tools themselves and in the evaluation process used. Automating documentation through artificial intelligence not only enhances code comprehension but also facilitates its maintenance, resulting in greater efficiency and cost reduction for organizations.