Enhancing Textual Emotion Recognition to Aid People with Disabilities Using Brown-Bear Optimization with Stacked Supervised Autoencoder Model

Emotional interaction plays a vital role in human communication and decision-making. Textual Emotion Recognition (TER), a key area in Natural Language Processing (NLP), focuses on automatically identifying emotions from textual data. TER has garnered considerable academic and commercial interest due to its potential applications. Recent advances in deep learning (DL) have significantly improved the accuracy of emotion recognition systems by effectively extracting complex features from text. This study introduces a novel model, Leveraging Brown-Bear Optimization and Word Embedding for Enhancing Textual Emotion Recognition (LBBOWE-ETER), designed to assist people with disabilities by optimizing TER. The process begins with comprehensive text preprocessing to clean and transform raw text data. The model employs GloVe for word embedding to capture semantic relationships effectively. For classification, a Stacked Supervised Autoencoder (SSAE) is developed, capable of deep feature extraction. To enhance performance, the Brown-Bear Optimization (BBO) algorithm fine-tunes the SSAE hyperparameters. Extensive experiments validate the proposed model’s superior performance on several evaluation metrics, demonstrating its robustness and practical relevance. The LBBOWE-ETER approach highlights the potential of integrating advanced word embedding and optimization techniques to advance TER systems, offering significant benefits for accessibility tools and emotion-aware applications.