An intelligent emotion recognition system for people with disabilities using stacked supervised autoencoder and brown-bear optimization strategy

Emotional interaction is a normal psychological fact in the daily life of humans. Accurate emotion recognition is the background of efficient human decision-making, communication, and interaction. With the evolution of artificial intelligence (AI) and big data, emotion recognition has now become a standard study activity. Textual language is the most standard human emotion carrier. In natural language processing (NLP), textual emotion recognition (TER) has now become a significant area because of its important academic and commercial potential. TER automatically identifies emotional conditions in textual languages. TER is an additional in-depth study beyond sentiment analysis and has attained great interest from the research area. With the progressive evolution of deep learning (DL) technologies, TER has received increased attention and has been considerably encouraged in recent times. The DL technologies had surpassed the confidence in physical feature extraction and attained good performance. In this manuscript, a Leveraging Brown-Bear Optimisation and Word Embedding Model for Enhancing Textual Emotion Recognition (LBBOWE-ETER) method is proposed to aid people with disabilities. Primarily, the text pre-processing phase involves several levels to clean and transform raw text information. Besides, the Glove method is used for the word embedding process. For the classification process, the LBBOWE-ETER method utilises the stacked supervised autoencoder (SSAE) model. Lastly, the brown-bear optimisation (BBO) model is utilised for the fine-tuning of SSAE hyperparameters. The experimental validation of the LBBOWE-ETER method portrayed a superior accuracy value of 98.95% over existing models under the Emotion Detection from Text dataset.