Identification of texts generated by AI (ChatGPT)

Abstract

This bachelor's final project aims to address the challenge of identifying texts generated by artificial intelligence (AI) systems. The project implements several classification models, including Classification Trees, Random Forests, Logistic Models, and Support Vector Machines to identify AI-generated texts. These models are trained on a dataset which consists of 160 texts of human and AI-generated texts, with the goal of accurately distinguishing them. The project also includes the implementation of a Shiny application, providing a user-friendly interface for text identification. Among the models evaluated, the Logistic Model achieves the highest accuracy, with 86%.