شيمائ علي

Machine-generated text detection is the process of identifying text that has been created by artificial intelligence models, to differentiate it from text that humans have written. Text generation technology has become increasingly important due to the potential misuse of text-generative models for creating fake news, reviews, or spam content that can deceive individuals. The objective of this study is to present a model that can be used to determine whether the tweet's author is a human or a robot. For this purpose, we employed a zero-shot pro‎mp‎t and pre-trained model and SBERT fine-tuned using different transformer models. Also, to extract the characteristics of the author's writing style, we have used the graph attention network and graph convolutional network models. According to the results, when the author's writing style is extracted with the graph convolutional network model, the best result is obtained with an accuracy value of 93.60%. Machine-generated text detection is crucial for protecting against the misuse of AI models. It helps to improve the accuracy in distinguishing between human-written and machine-generated content, thus enhancing trust in online platforms

تشخيص متن توليد شده توسط ماشين , اعلام ضربه صفر

Machine-generated text detection , Zero-shot pro‎mp‎t

Sheyma Ali

Dr. Behrooz Minaei