ChatGpt-generated modifications on human generated TESOL abstracts by Vietnamese researchers

Abstract

As a potential writing assistant, OpenAI's ChatGPT has been explored for its capability in enhancing academic writing skills, especially through its ability to edit or polish human-generated versions. This study, therefore, aimed to investigate the differences between human-generated TESOL research article abstracts written by Vietnamese authors and their ChatGPT-edited versions in terms of move structures and linguistic features through understanding the textual modifications generated by ChatGPT. The findings from textual investigation involving the analyses of abstract types following Lorés (2004), moves based on Hyland's (2000) framework, and key linguistic features of 50 original abstracts and 50 corresponding ChatGPT-edited ones indicated significant modifications in structural, lexical, sentential and tense dimensions of the two sets. Besides, interviews with two experienced Vietnamese abstract and manuscript reviewers were conducted to elicit their opinions on the quality of the modifications identified. Both our textual findings and interview insights complemented each other that ChatGPT-generated modifications improved clarity, readability and informativeness of the original abstracts. This paper ends with proposed pedagogical implications drawn from our findings to help not only Vietnamese researchers in TESOL and other related fields understand and appreciate such modifications, which can support them to produce research article abstracts more effectively.