Emotion Recognition System to Suggest and Respond to Vietnamese Using PhoBert Model

The article proposes an emotion recognition system to suggest and automatically respond in conversations to increase user experience and give immediate feedback. We propose to build an emotion recognition system including six types of emotions (disgust, pleasure, sad, fear, anger, and others) using the PhoBERT model. Besides, we use the dataset built in the article [1] to compare and demonstrate the improvement from the proposal model. The input data is text. The output is the emotional label corresponding to the input. The system has an accuracy of up to 99.8% with the training dataset and 96.4% with the testing dataset. This result is an improvement 15% comparedto [1] showing that we can build a more accurate emotion recognition system for users and be applicable in practice.