EMOTION DETECTION VIA BERT-BASED DEEP LEARNING APPROACHES IN NATURAL LANGUAGE PROCESSING

This study focuses on emotion detection using BERT-based deep learning approaches in the field of natural language processing (NLP). Unlike traditional methods, the BERT model exhibits superior performance in sentiment analysis with its ability to produce bidirectional contextual representations. In the study, a dataset consisting of social media posts written in Sundanese language was used and four main emotional states (anger, enthusiasm, anxiety, and melancholy) were classified. In the data preprocessing stage, the special characteristics of the language and the informal structure of the social media language were taken into account. The performance of the BERT model was evaluated using metrics such as accuracy, precision, sensitivity, and F1 score and compared with other methods. Experimental results show that BERT-based models provide high accuracy and reliability in sentiment detection tasks. In addition, the contextual understanding capability of the BERT model provided a significant advantage in overcoming previously encountered classification challenges. The findings show that BERT-based sentiment detection models can be effectively used in various applications such as social media analysis, customer feedback evaluation, and brand reputation management. This study provides an important contribution to the development of more effective and reliable methods for sentiment analysis in the field of NLP.