Sharif-MGTD at SemEval-2024 Task 8: A Transformer-Based Approach to Detect Machine Generated Text

TL;DR

This paper presents a transformer-based approach using fine-tuned RoBERTa to detect machine-generated text, achieving competitive accuracy in a SemEval-2024 task with limited hardware resources.

Contribution

The study demonstrates the effectiveness of fine-tuning RoBERTa for MGT detection within resource constraints, focusing on monolingual English texts in a competitive setting.

Findings

Achieved 78.9% accuracy on test data

Ranked 57th among participants

Effective in identifying human-written text

Abstract

Detecting Machine-Generated Text (MGT) has emerged as a significant area of study within Natural Language Processing. While language models generate text, they often leave discernible traces, which can be scrutinized using either traditional feature-based methods or more advanced neural language models. In this research, we explore the effectiveness of fine-tuning a RoBERTa-base transformer, a powerful neural architecture, to address MGT detection as a binary classification task. Focusing specifically on Subtask A (Monolingual-English) within the SemEval-2024 competition framework, our proposed system achieves an accuracy of 78.9% on the test dataset, positioning us at 57th among participants. Our study addresses this challenge while considering the limited hardware resources, resulting in a system that excels at identifying human-written texts but encounters challenges in accurately discerning MGTs.