Attention Convolutional Neural Network for Sentiment Analysis on ChatGPT

Sentiment analysis is a technique that is able to analyze reviews, sentiment, people’s behavior and emotions towards entities such as services, products, organizations, events and social media. Several sentiment analysis approaches have been carried out, one of which uses deep learning such as Convolutional Neural Network (CNN). This research implements the Attentive model using 3 models, namely Attention network, Multi-head Attention and Convolutional Block Attention Module (CBAM). Attentive mechanisms are used to capture important words so that the model can understand important information. Combining Attentive and CNN models to find optimal performance for each model can be done by hyperparameters tuning settings. From the results of selecting the best hyperparameters, it proves that the Attentive mechanism model is able to increase the performance of the baseline model, each accuracy value that increases is Attention network 0.37%, Multi-head Attention 0.43%, and CBAM 0.21%. The highest improvement value was obtained by Multi-head Attention.