Automated Scientific Article Generator System Based on GPT Algorithm

Abstract

Composing scientific articles and synthesizing relevant literature are often time-consuming and challenging tasks for researchers. This study developed an automated scientific article generator system that leverages advanced Artificial Intelligence capabilities to address these inefficiencies. The proposed system integrates the OpenAI API using the GPT algorithm to construct natural language generation with Web Scraping techniques, specifically targeting academic databases such as IEEE Xplore, to dynamically retrieve and incorporate up-to-date scholarly references. The system is designed to streamline the article writing process by generating cohesive, structured text based on user-defined topics and seamlessly embedding pertinent citations. The performance of the generated articles was rigorously evaluated using quantitative metrics: ROUGE (for lexical overlap) and BERTScore (for semantic similarity) against reference texts. Empirical results are highly promising: the system achieved a BERTScore F1-Score of 84.46%, demonstrating superior semantic correspondence and contextual relevance while extracting critical information from source texts. This proposed technique can be a potential solution to enhance writing efficiency and support academic documentation.