Application of the Deep Neural Networks Model in Analyzing ChatGPT Application Sentiment

Abstract

AI has been able to intelligently mimic human behavior and has been applied in various contexts, including healthcare for more efficient patient care. One of the prominent trends in AI is advanced language models such as ChatGPT developed by OpenAI. The effectiveness of ChatGPT in finding and fixing bugs in computer code is a subject of debate, depending on the task, training data, and system design. The popularity of social media platforms, particularly Twitter, as a data source for text analysis has increased interest in sentiment analysis. This study explores sentiment towards the ChatGPT application using a dataset of 50,000 tweets. Sentiment analysis is performed using a deep neural network (DNNs) approach to achieve optimal accuracy. Deep learning models are known to have high predictivity and efficient training time. Through this experiment, we aim to gain insight into how the public views ChatGPT in three sentiment categories: positive, negative, and neutral. DNN (Deep Neural Network) is proposed because of its good performance and can shorten the amount of training time. The results with the model used in this study, namely CNN and LSTM both achieve an accuracy value of more than 90%: Where CNN obtains an accuracy value of 91.12% and LSTM obtains an accuracy of 90.84%.