Year: 2025
Type: dissertation
Abstract: Contemporary software development faces significant challenges related to insufficient, fragmented, or outdated technical documentation. Developers frequently need to maintain complex mental models of systems while working, resulting in context loss and trial-and-error approaches. This work presented the creation of an intelligent chatbot designed to assist developers in understanding and navigating through complex systems. Using advanced natural language models, the developed chatbot is capable of processing, indexing, and querying existing technical documentation, answering developers’ questions about flows, business rules, and internal system structures. The methodology included the development of a web interface integrated with the Retrieval-Augmented Generation (RAG) architecture and a vector database. The results obtained demonstrated a precision of 85% and an F1-Score of 0.83 in information retrieval. Validation with developers indicated an overall satisfaction of 4.7/5, highlighting ease of use, although the average response time (25.2s) was identified as a limitation. It is concluded that this approach reduces the time needed to understand complex systems and facilitates the integration of new members into development teams.
Author: Mayer, Oliver Lohann
Language: Portuguese
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7155674423