ChatGPT Chats Decoded: Uncovering Prompt Patterns for Superior Solutions in Software Development Lifecycle

The advent of Large Language Models (LLMs) like ChatGPT has markedly transformed software development, aiding tasks from code generation to issue resolution with their human-like text generation. Nevertheless, the effectiveness of these models greatly depends on the nature of the prompts given by developers. Therefore, this study delves into the DevGPT dataset, a rich collection of developer-ChatGPT dialogues, to unearth the patterns in prompts that lead to effective problem resolutions. The underlying motivation for this research is to enhance the collaboration between human developers and AI tools, thereby improving productivity and problem-solving efficacy in software development. Utilizing a combination of textual analysis and data-driven approaches, this paper seeks to identify the attributes of prompts that are associated with successful interactions, providing crucial insights for the strategic employment of ChatGPT in software engineering environments.CCS CONCEPTS•Information systems → Data mining.