Deep Learning-Based Detection of Machine-Generated Tweets with FastText Word Embeddings

Abstract

The rapid growth of social media has made it easier for information to spread quickly, but it has also increased the circulation of machine-generated and misleading content. Advanced language models can now produce tweets that closely resemble human writing, making it difficult to identify fake content through manual inspection. This project presents a deep learning approach for detecting machinegenerated tweets using FastText word embeddings and a Convolutional Neural Network (CNN). The collected tweet dataset is first preprocessed by removing unwanted characters, stop words, and noise to improve text quality. FastText is then used to convert the cleaned text into meaningful vector representations that preserve semantic information. These embeddings are provided as input to the CNN model for classification. The proposed approach effectively distinguishes human-written tweets from machine-generated ones and demonstrates better performance than conventional machine learning methods. The developed system can support social media platforms in reducing the spread of automated misinformation and improving the reliability of online communication.