What is it about?
The article documents our method to distinguish author written and AI generated abstracts of scientific scholarly texts. We show that suitable semantic and pragmatic features of the text are effective discriminators, using classical machine learning algorithms.
Featured Image
Why is it important?
Use of AI in generating scientific scholarly text is a worrisome trend, since it negatively impacts the growth of science. With strengthening and continuous advancements in LLMs, the prevalence of AI generated text in is likely to increase in school and college assignments, in addition to scientific writing. This warrants constant progress in automated methods to identify AI generated text , particularly in scenarios where it is unethical.
Perspectives
In this article, we were able to achieve reasonable accuracy with a small training set. The credit goes to the selected semantic and pragmatic features. features I hope this article will bring to spotlight the importance of linguistic features, which probably LLMs can't learn.
Vasudha Bhatnagar
University of Delhi
Read the Original
This page is a summary of: Deep dive into language traits of AI-generated Abstracts, January 2024, ACM (Association for Computing Machinery),
DOI: 10.1145/3632410.3632471.
You can read the full text:
Contributors
The following have contributed to this page
