Empowering people with intellectual disabilities using integrated deep learning architecture driven enhanced...

Area of Science:

• Artificial Intelligence
• Natural Language Processing
• Computational Linguistics

Background:

• Emotion recognition is crucial in psychology, healthcare, and human-computer interaction (HCI).
• Conventional methods for textual emotion recognition (TER) often rely on limited data and have reliability issues.
• Deep learning (DL) advancements have spurred significant progress in TER.

Purpose of the Study:

• To develop an advanced deep learning system for accurate text emotion recognition.
• To enhance communication accessibility for people with disabilities through improved TER.
• To introduce the Intelligent Emotion Recognition from Text Using a Hybrid Deep Learning Model and Word Embedding Process (IERT-HDLMWEP) model.

Main Methods:

• Text pre-processing to reduce dimensionality and prepare data for analysis.
• Hybrid feature representation integrating Word2Vec, TF-IDF weighting, and Part-of-Speech features.
• Classification using a hybrid Convolutional Neural Network (CNN) and Bidirectional Gated Recurrent Unit (BiGRU) with an attention mechanism (C-BiG-A).

Main Results:

• The IERT-HDLMWEP model demonstrated superior performance in emotion detection from text datasets.
• Empirical results confirmed the effectiveness of the proposed hybrid deep learning approach.
• The methodology showed significant improvements over existing TER techniques.

Conclusions:

• The IERT-HDLMWEP model offers a robust and accurate solution for textual emotion recognition.
• This approach has the potential to significantly aid individuals with communication disabilities.
• The study highlights the efficacy of hybrid deep learning architectures in advancing NLP tasks.