Integration of corpus linguistics and deep learning techniques for enhanced semantic-driven emotion detection on...

Area of Science:

• Natural Language Processing
• Computational Linguistics
• Machine Learning

Background:

• Corpus linguistics offers a framework for analyzing language structures and patterns.
• Text-based emotion recognition (TER) utilizes linguistic features from annotated corpora to train machine learning (ML) models.
• Integrating corpus-based and data-driven methods enhances emotion recognition accuracy for applications like sentiment analysis and mental health assessment.

Purpose of the Study:

• To propose an effective textual emotion recognition and analysis approach to improve sentiment understanding.
• To develop an Emotion Detection in Text via Integrated Word Vector Representation and Multi-Model Deep Neural Networks (EDTIWVR-MDNN) method.

Main Methods:

• Text pre-processing for relevant data extraction.
• Word embedding using a fusion of Term Frequency-Inverse Document Frequency (TF-IDF), Bidirectional Encoder Representations from Transformers (BERT), and GloVe.
• Classification using a hybrid Attention Mechanism-based Temporal Convolutional Network and Bi-directional Gated Recurrent Unit (AM-T-BiG) model.

Main Results:

• The EDTIWVR-MDNN methodology achieved a superior accuracy of 99.26% on the Emotion Detection from Text dataset.
• Experimental evaluation demonstrated significant improvements over existing approaches in emotion recognition tasks.
• The integrated word vector representation and multi-model deep neural networks effectively capture complex emotional expressions in text.

Conclusions:

• The proposed EDTIWVR-MDNN approach significantly enhances the accuracy and depth of textual emotion recognition.
• This method offers valuable improvements for sentiment analysis, social media monitoring, mental health assessment, and human-computer interaction.
• The fusion of advanced NLP techniques and deep learning models provides a robust framework for understanding sentiment in natural language.