Comparative Evaluation of CNN-LSTM Model for Emotion Detection in Indonesian Text

Abstract

This study developed an emotion detection model for Indonesian text using a dataset from prior research. The data was refined through multiple pre-processing steps before applying CNN-LSTM machine learning techniques. Comparative analysis indicated the model achieved 58% accuracy, lower than baseline methods. The results imply need for larger annotated corpora, improved text normalization, and integration with state-of-the-art deep learning approaches to enhance performance for Indonesian emotion detection.