Related Experiment Video
Updated: Jun 6, 2025
10:04
Construction of an Improved Multi-Tetrode Hyperdrive for Large-Scale Neural Recording in Behaving Rats
Published on: May 9, 2018
11.3K
Using generative adversarial network to improve the accuracy of detecting AI-generated tweets
1School of Humanities and Law, Zhengzhou Shengda University, Zhengzhou, 451191, Henan, China. huiyang0909@126.com.
Scientific Reports
|November 26, 2024
Summary
This study introduces a novel AI-driven method for detecting AI-generated text, achieving 99.60% accuracy by using generative models and ensemble learning for precise text attribute characterization.
Area of Science:
- Computer Science
- Artificial Intelligence
- Natural Language Processing
Background:
- The proliferation of AI-generated text necessitates robust detection methods.
- Existing machine learning techniques require enhancement for accurate AI text identification.
Purpose of the Study:
- To develop a novel approach for improving the accuracy of AI-generated text detection.
- To leverage generative AI capabilities and ensemble learning for precise text attribute characterization.
Main Methods:
- Text preprocessing including noise removal, tokenization, stop-word removal, and normalization.
- Feature engineering representing text as matrices capturing word correlations and weights.
- Generative Adversarial Network (GAN)-based feature extraction.
- Weighted Random Forest (RF) model for final text classification.
Main Results:
- The proposed methodology achieved an average accuracy of 99.60% in distinguishing human-written from AI-generated text.
- Demonstrated a 1.5% improvement in accuracy compared to existing methods.
- Successfully characterized attributes specific to AI-generated content.
Conclusions:
- The novel approach effectively enhances AI-generated text detection accuracy.
- Combining generative AI with ensemble methods offers a powerful solution for content authenticity.
- The method provides a significant advancement in identifying AI-created content.
