Detecting ChatGPT-Generated Text and Human writen Text Using Deep Learning

Recently, the identification of text if it is written by human or ChatGPT-generated has become a hot research topic. The current work presents a Memory Recurrent Neural Network (LSTMRNN) model to detect both human as well as ChatGPT-generated text. The purpose of the proposed LSTMRNN method is to investigate the model’s decision and detect the presence of any particular pattern. In addition to this, the LSTMRNN technique focuses on designing Term Frequency–Inverse Document Frequency (TF-IDF) word embedding, and count vectorizers for the feature extraction process. The simulation performance of the proposed LSTMRNN technique was investigated on benchmark databases, and the outcome demonstrated the advantage of the LSTMRNN system over other recent methods (SVM, Basic Deep learning, Convolution Model (CNN) LSTM Model (RNN)). The Data Set has 10,000 records. [5,000 records are taken from ChatGPT generated Arabic Dataset CIDAR and other 5000 records taken from human-written articles MNAD (Moroccan News Articles) Dataset. both are merged to get 10,000 records. This bench marked data set is used to train - BIDIRECTION-LSTMRNN modal. Note: 1. SVM model source code could be found in NLP_SVM.py and results in lstm-rnn-mdl-result.txt 2. Basic DeepLearning model source code could be found in NLP_DL2.py and results in basic_dl_mdl_result.txt 3. CNN model source code could be found in NLP_CNN2.py and results in cnn_mdl_result.txt 4. DataSet with 10000 records could be found in human_chatgpt_genarated_dataset.csv 5. All these files are available at: https://github.com/bbaktech/Differentiating_ChatGPT_AND_Humans.git