An Empirical Study of ChatGPT-related projects on GitHub

As ChatGPT possesses powerful capabilities in natural language processing and code analysis, it has received widespread attention since its launch. Developers have applied its powerful capabilities to various domains through software projects which are hosted on the largest open-source platform (GitHub) worldwide. Simultaneously, these projects have triggered extensive discussions. In order to comprehend the research content of these projects and understand the potential requirements discussed, we collected ChatGPT-related projects from the GitHub platform and utilized the LDA topic model to identify the discussion topics. Specifically, we selected 200 projects, categorizing them into three primary categories through analyzing their descriptions: ChatGPT implementation & training, ChatGPT application, ChatGPT improvement & extension. Subsequently, we employed the LDA topic model to identify 10 topics from issue texts, and compared the distribution and evolution trend of the discovered topics within the three primary project categories. Our observations include (1) The number of projects growing in a single month for the three primary project categories are closely associated with the development of ChatGPT. (2) There exist significant variations in the popularity of each topic for the three primary project categories. (3) The monthly changes in the absolute impact of each topic for the three primary project categories are diverse, which is often closely associated with the variation in the number of projects owned by that category. (4) With the passage of time, the relative impact of each topic exhibits different development trends in the three primary project categories. Based on these findings, we discuss implications for developers and users.