Between December 5th, 2022 and June 7th, 2023, a total of 1313 scholarly documents on ChatGPT, including papers, preprints, and editorial material, have been published.
According to data gathered by Google Scholar, the first scholarly document published was "Performing Effective Research Using ChatGPT," featured in the November-December 2022 issue of the Indian Journal of Computer Science.
As an artificial intelligence-based language model, ChatGPT has generated a lot of interest in the scientific community. In this analysis, an overview of the current scientific output on ChatGPT is provided.
Throughout the period, the average number of new documents produced per day was 7.4, with a peak of 11.8 documents per day in May.
Publications have focused on a variety of topics related to ChatGPT, including:
1. Model improvements: much of the research has focused on improvements to ChatGPT's architecture and training techniques to improve its accuracy and effectiveness.
2. ChatGPT applications: there are also a large number of publications exploring the applications of ChatGPT in areas such as automatic translation, natural language processing, sentiment analysis, text generation, and reading comprehension.
3. Ethical and social aspects: some publications have analyzed the potential negative effects of ChatGPT usage on society, including gender bias and possible manipulation of public opinion.
4. Evaluation and comparison: there are also a number of publications that have focused on evaluating and comparing ChatGPT with other language models and artificial intelligence systems.
Overall, the scientific output on ChatGPT reflects its wide application and importance as a language model in the field of artificial intelligence. Current research focuses on improving the model and exploring new applications, as well as considering the ethical and social aspects of its use.
The present study systematically reviews the scientific production associated with ChatGPT.
To conduct the study, the data is searched on Google Scholar using the keywords (ChatGPT OR "Chat GPT").
Studies are restricted to those published from November 2022 (Beginning of the ChatGPT by OpenAI) to the date of the latest search.
No language restrictions are applied.
Types of study included
Original articles, working papers, cross sectional, cohort studies, case-series, case reports, conference abstracts, reviews, editorials, pre-prints, unpublished data and studies involving an analysis of ChatGPT results are eligible for inclusion.
News, comments, expert opinions, blogs and laboratory features of individual cases are excluded.
Selection of studies
Those references that do not include our unit of analysis in the title, abstract or keyword are excluded. A total of 1313 scholarly documents on ChatGPT were identified. 40% of these papers have received at least one citation. 48% of the publications are preprints or early access papers.
We've incorporated a tag cloud visualization that displays the most common terms extracted from the papers titles. Among the most frequently occurring themes in the scholarly publications are words like chatgpt (799); language (262); models (231); large (185); ai (131). The graph is interactive, and by clicking on it, you can access the entire corpus and experiment with it in the Voyant Tools application.