Scientific Productivity on ChatGPT: A Bibliometric Analysis

Abstract

Introduction: The discipline of Natural Language Processing (NLP) has experienced unprecedented advancements in recent years. Among these, OpenAI’s ChatGPT has emerged as a frontrunner, captivating students, researchers, and enthusiasts alike. As ChatGPT advances further, a need has arisen to judge, assess, and understand the pattern and trajectory of scholarly contributions in the form of a ‘Bibliometric Method’.

Motive: This article takes a bibliometric approach on 2302 scholarly publications related to ChatGPT from its inceptions year to 2023. It performed author productivity, citation analysis, keyword co-occurrence, and productivity of journals and authors. It also performed various collaborative measures as well as Lotka’s law of scientific productivity.

Methodology: Quantitative bibliometric analysis was chosen as the methodology for this research. Scopus was picked out to be the database to collect data. 2302 documents fulfilled the search query and thus, were chosen as the dataset for this research. Data refining and all the related works were performed in MS Excel and. Vos-viewer and biblioshiny ware used to visualize the data.

Findings: after the analysis, it was found that most of the documents written over ChatGPT were articles, authors preferred collaboration over individual works, keywords i.e., artificial intelligence, large language models and ChatBot co-occur distinctively with ChatGPT, USA is the top productive country whereas Journal of Biomedical Engineering published most work over ChatGPT. It was also observed that the collaborative pattern of authors does fulfil ‘Lotka’s law of scientific productivity’.

Originality: As ChatGPT is comparatively a recently emerging concept, not a lot of bibliometric research has been performed on it. Thus, this research is one of the pioneers in ChatGPT-related bibliometric analysis and wishes to pave the way for future research.