Social media opinion manipulators now have more tools at their disposal because to recent developments in natural language processing. access to an additional tool. Furthermore, due to advancements in language modelling, deep neural models now possess greater generative skills, improving their capacity to produce content. Because of in recent years, text-generative models have grown in efficacy, which allows attackers to take use of these incredible skills to fortify social bots and create convincing deep fake posts that sway public opinion. Addressing this issue requires trustworthy and precise methods for detecting deep fake social media postings must be developed. Because of this, research on recognizing computer-generated content on social media networking sites like Twitter is still underway. This work classifies uses Twitter when either human- or bot-generated using word embedding’s and a rudimentary deep learning model using the publicly available Twee fake dataset. Using Fast Text word embedding’s, using a standard design for a CNN is created to identify deep fake tweets. Many machine learning models were used as reference approaches in this study to show the improved efficacy of the recommended methodology. These baseline techniques incorporated Fast Text, Fast Text sub word embedding’s, Term Frequency, and Term Frequency-Inverse Document Frequency. Furthermore, the advantages and effectiveness of the proposed approach are emphasized in comparison with other deep learning models, namely CNN-LSTM Together with LSTM systems, in successfully completing what needs doing. The experiment the results show that the convolutional neural network is suitable for accurately identifying twitter data with a 93% accuracy rate when combined with Fast Text embedding’s.

Text categorization, deep fake, machine learning, machine-generated text, and machine learning.